Skip to main content

Full text of "mit :: ai :: aim :: AITR-993"

See other formats


'*i^Uitoo 



A GOVERNMENT-BINDING BASED PARSER 

FOR WARLPIRl, 
A FREE- WORD ORDER LANGUAGE 

by 

Mi[:h£icl B. KbeIiIceI 



CEigijj^eriiig and ^'gE[i|>Mlrr SriFnrr on Jajiu^o ^U, IQ47 m patIlbI EuHiHiiiecLL iillhe 
[-^<|pir4iirn1-s Fnr iJi" rl^girF nP Matter af Science. 

ft} MaiAAchuiell^ Iiial]lLi1« of 'IccbnoLi^g} l&hT 



Tbii? report dc^rihcs r^^carch done at tti? ArtLfic]al 1ntel]j|^ihce LFihoiAt^^ i^r 
tbi- MftiNft^liMH^cL^ Ici'inriritr cf TtirfanolfiKTr SiijapnirL Tor rfir ]^b4r^43r¥'s artifiiiiaJ 
int^lLgcRci? rr^'ar^h Lb provided in p&rl h^ th? Advanrud Hesearcti Pi^je^t^ A^h^h^^ 
ig ih« L>eparti]»nt Chf JJefeme fUAJfPA) uiictcr Offlrn^ L^f N^i/.d Rtr^r^rdi (ONfl.) 
concr^t N[>UD 14-^0- K-UL 24: Ln part bj d. ilienerEil Electric 1eJkiwsb][7; ^3) p^L bf 
DAttPA under ONk rontract Nflfffi I l-Rfl-< '- RFiflFi ; and in p^rt h^- ^ NriUunal ^ciFnci? 



Abstract 



Fnie-wctrd cvderlnji^uase^ It^ive |r>rtg prhscd ^ianificant prubletp].-^ fof »tAnd?irJ E^^^'Off 
J^l^nihiiis- ThiE theBLs presfnli mi Lrtll^krtn^trd |3ir**rT hflsed cm Gov^rnilWnl Bir- 
ding iPB) thn^rtry. Tar 4; p^irfkialcir Jree-'Acird urcliir LyifciiTi^CT W^irlj^irLn an jthoriRmnl 
Inn^uAgf^ oTcfntrAl Auvirn-iUn. 

The words in asentfJI^f H>f ^ iS*tii wm-tl order langunRt iTlJ*y ^w^p Aljnjut jrlativeJy 
frerlv with [Ltcle etfect on m^ftiihng: tbn^ |>i?riuijtatijnL l-F a wntH^nc^ lllJ^n csMntiaJiy 
Lie aan^e rhmgr U is as^uiued that tli|& imillarily in i]iT'aning ll dij#crly reHpfced Iji 
the iiyiitai- The |>ar*tr prcwnred her? properly prorrNi^rs fir^ word order Ij^cjut^^ 
\^ hs^ign? the 3Ai]]e -^yntJirtK -:1 ructun: to the perih^hLtAlicpns oF a ilnglc scntirace. 
The pars^ ^St> h^Nndlrft fixed word cirdi?r, a.^ iveU ?w oLher phwioniena. On the virifl 
presented ber^s tlhti^ |$ ny ^lach thin^AS a " confi^urAl 11^1141!" or '^noii-configiirat]ona|^ 
langUagJ^, Bachcr. there lb a sp^cEfiMiP of lETTii^ua'^i^B thai aie nK-re nf lr'-^ urrlrrrj. 

The op^rMlon r>f this p^u^lni; syslenk h qnitt^ JjlTcrcnt in cLajActer front lh*t 
of ii]i:irG Lradhional riiJe-hA^fHl tHEirsin^ Bystenia, e_^J_^ frhnt^KlTrrr |>H";r5cr*- In thij 
^y■^trl[L, parking is cnrj-bed Lnit y]^ thr rrjnstmction of Iwlj d|(f^r^tL( siru^LnresT one 
encodme pr?i7tH-li^n cc inTorEnatioD and onp entc^<liria hirrarchLral jnforciianon- ThJ^ 
hiparlLlerepreHt'jltauoJils thrlicy tohandJiriRhoThfrrr ?;ni] fiKcd order ph^nfrinf-na. 
This ihwL^ firat prest^fllM ^n otrrvi^w cf the poj-Lli^n of Warl|>[ri Ihat tan he 
pi^rsril. FoLloiiine this Ei a d^ifriprimn \y[ th^ tin^uistLc (htorj t>rq w^hLch Ihc parser 
Ik based- Til* chat>lH?r afttr that iJe=:trtii^i. tli£ representations ajid alRc^rhtii^n^ of 
Ih^ parser. Jn conciusi^ifcT fbr par^T h coinpartfiJ Us r^l^J^d *i?rk. The appendix 
ft>tlt*Jns a luhstantial iJsL of tfflt rMcfl— h^ih ffraiiuiiaUcal and nn5r4wiLnialical— 
that the parser iu^ iKtu^dlr processed. 

Thr>^i< Snpefh-iiEif' Rd^bn^rt C Berwick 

Title: flwocialc Professor of CutcLpulcr Sc^irnce and Lnglneeflflj 



Acknowledgments 



I whb to thank Eik^ fheai^ odviNorn Hoh^^rL HerNvirk. fnr bif KF[prul advice and crltJ- 
tkniiiH- J idjo whh \u chank Haj> Laughren for i\rT iiP^Lfiitlicin nn Wiirlpiri. withuut 
wiudi I hVLjulit iiMt Jiti^t IjH^n o^pIc Cq crratc this i^arser. 

TUff theaj^ U dfKllrntcd to ni) Jiucnts, E^A and ^tiBlhy KA.>hLi'tr 



Contents 



Abstract f 

Ackikci^lrrigmtnrs j 

1 Int roduEtion g 

lA A WarlijIrL J'rLDk&r 10 

1-Ll Tlir MEJnr Parl^ of .^pt^h , IJ 

1.1.2 fast fkuses ?| 

Lrl-3 Agiccmert . _ ^ 22 

1.L.4 Afciiilirvy PCKirLrTjiinS , , y^ 

InlrS ArUunirnt rdenfLfLtJitiLiCi - n r + iA 

l.LB Null Aiti^hcifii , ?S 

I-? TIk Mtlhcdologv ul Natural (.'onipuuCL4jii , , 35 

l-:i The Ahslract Ci^a^puUHk^cLnl Thror> + , . . 2S 

1.4 CoTtUflR AttCnclioDs ^ 32 

3 Thti Tj liLg ^119 r Lc I'heory gj 

2.1 PrtceiitiKr StrncLurp .....__, 37 

2.1.1 Auriliftfy 1'utlipD^ltii>i] + h , . r ^l 

? L-2 AdtiliAi} Po^Jtioniiifi ^ ^ , ^2 

2.2 Syntai . , ^ ...-_., , , . . 43 

2,2.1 X thijurj _ ^ 44 

2.a.2 L-atruftUfE^ , H H , r 1 . . . 4B 

2-2rJ D-ttructure _ ^ - ^ ^ , * ^ , 51 

2.2.4 S-striiL'[urH^ 1^2 

i!-2 5 Till- Mappi^a BeLwHi> i-blrdcturcand D-atruttHrt . r . . . . B5 

5.;! Thr LcxLci^n + ,, *,, 57 

7.3.1 1 hn^ T^xkal Entry . , ^ , . , . . 5fl 

2 3 2 L-zjica] Rulei . . , 59 

2,4 Sn^mantLC InierpreifLLJLhri ^ , h,r. fill 

2.4.1 Arfluiihjjit iLlmtiJif^Aiicin .....__,,, n . . . fiD 

2.4_2 Nhkll AuJtilinry (TompDnEnt^ + r , , DL 

2-4-3 Nail Anaphorfl . . n ^ jl 

2,4r4 ScniATitlc WEll-ryrTtl^lcLnH * - , Gl 



Rcpre^entatEoii and Algi^k-lthm 44 

3.1 IJH^j^r^^E'Et^^ciijcL X . X _ _ f]^ 

.l-U Preccdenc4* SlriiL'hirt atjij SyML'M'lic Stnictur** - l _ h f]?; 

^rl-2 Thn^ LrKirrtn ^ h i- , , g^ 

J? AlgorLttiin . . , l . . . . _ h , , 73 

J.;!J The ['a F^ATser 74 

3Jj3 Thr SynUc^ck ["flrncr . ^ - , , . ST 

3-2,3 StiilWitk [nLLTf>n'rEitm[i ^ ^ . _ ^ 9| 

3.3 P^kriing [ti« Sftiupl4i Senitttt*^ , r , 91 

CcHiKliisioH M 

4rl Olbrr Cn'U]ii]iAiLCFU TniLiLework-i _ _ h * ^ h _ ^ , , - 9Q 

i-LL ID/LF tlf^tcLiLHu- J . . . _ , lOD 

4.1.2 LMicLif-f'uiiL'rlfKi.il tin^ttulinr ^ lOj} 

4.I-J Tr?p- A JjoiDliig ti r4Hj] imAT 1[J2 

d-J ^hE^i-koniing^ and Fufun- Worl , , |03 

12J Mort- Warlplri ....-._ ^ ., , . . loa 

4-i^.2 Other LEuifiuagiiB h*____ *_ + rrx lOT 

Teat Ctiti lOi 

A.L InhpLen^entatid^ N^t-ff *_* 1^9 

A. 2 Trflts (!ascs ^ , , r - ■ - - - lfl9 

A-J-l Vn^rfi^ Striiia . . h - , - . ^ llO 

A. 1^-2 InHt^L-rpfl \VHj^ .. 1 ...... j 1 ... J . [J3 

A.2r3 Nciin C'on]pcisiticir; _ ^ , , , , ^ ]IH 

A. 2,4 Vj?rb rom[j[H]tLC]n , II7 

A-?-5 Auj^lJifLtt C'i>m|3«itLcin - - - n ._ + .., IJd 

A_?_t (:£intinm>U^ t'cwr Phrasw . . H X H + + + , , Vii 

A.2.T Aiijdiiaj-y Pcjsinrhning , , ^ ^ _ 133 

A.2.K FrtiP PKjELse Onlar IHD 

A-2.9 Arpuinent iHirnlilitflti™ . . . . . . + |4l 

A.3.10 Null Aiajilmij Cci[lipMicii:L^ H3 

Ar2,ll Null AnaphorFi _, *-**_,, IM 

Ar2-12 Toti MrinV Argiimcntfi ^ ^ J&7 

A/JlLZ TflRe M&rfcrcig I5g 

A-?. 14 Aiiiiliar^ Baatj AgrMcjuEat * + _,rx I6l 

A.ZIS Nnmln^l AiyrrnicnT h , , , r - , lft4 



Chapter 1 
Introduction 



Thli thf^li pieaentB & solution ftjr the |>r^vioa5L}' uflsoU^iJ prolil^uj E^fpaning £rpc* 
WL>nJ €jfd€r]iiLiRii?iseBJ^ In Lhn^Bc [?inaiagp^T Uir V^itttls in ?l s^nl^nn^^ ni^y-^^^p^hoiil 
rclati^rl^v frrti|y wich LU^Ff rFfe^t i^hn thh^j^n^nj;: the pernkiitjatLCinE ofa senlence nieBn 
c^scntiriL^ ilhr vaic^ thing. U Is Fi^siLCh^eJ that thE sLiuJJarity m nicaninjf h directly 
reflected tn the fynlax. So^ ^ parwr th?;c pti>pH^rly prcire-EHes rrttf word i>rdpf Fh^^L 
assign the samt flyn^EKtJf StrnttLirt cc? the periiiLiLatbons of a sentence. The parser 
aJsO haiLrllr^ fw^ft word orders as wcJ eu (»th?r synlJirlir phcjiT?mrn^. 

ffntil r«ent]y, niEin^ m^tuml- iRnftU^gt^ pjtf^pf^ havp hten dtibh^ned ari>und cojit- 
piitAtinnrilLv ^Mfni^tliy FrHrn^li^niH, such a^ t'LHLTe 1 1- Erce graiiiihiEirE, rhat hftie littJe 
UnguhtLr n>iiiirUlic)ki. iii d^x^n Uie^e imi^^rt h^^e worked cQir^rUy on Irtit n Eim- 
iL^d subset ChJ riAtLiTal ii4 terEuices. Hov^Yrr, th«^ biav^ ^nv^ ^t Lh^ir r^LLlU qtiitti 
quickly^ The throry i^f p^r^mg pr^^^nti^rF jji this thp.'iJSL cm th» oLher haDd^ is has4^cl 
on one cujTcrit linguislh (he^^ryr llie r^huj! 3b that Lite LnLplanu^nLeJ parser OLilpnts 
linguist Icnily inruilagFul ^tcucLures corresponding to the inl^ut s^ntn^nc^. Thr nhaia 
hypcth^sJs iifrc i^ tiuit w» are mor? liki^ly to arrive at ii «i]^ C4^> sful p^r^r \i^r bA«n 
]t on liD^ui'itJC thpoT?, cathE-r than rnn ^itniput^tEi^il^^l i:^n^l(leTAt]on& nlone- 

The }:Arsinti niotfcl ciiUSt, i\i foiafi* h»i: tesleil lid a natural language. Warlpiri, 
AiL ala^jj-tf^irihaJ iaa^Liage fro[i3 centrFkl AuElraliai iv^hH th^sril lan^r^Ut^^ It i^ prrlL?ips thf* 
paradifiinaLLC natural lan^Uriifc- e^hihilinj^ frtrr Wi>ri3 i^rdeF- On a j^iore practical Itftl, 
WEirlpirihn':qL r^lntive |y i^iM^ple dynt^Landalairly sntall lejkon, which mak^fi lor an 
rn^ipT i^h ^jf prudiicinR a parser that handles an a|a|a]TT3rtblr suhsrt r>rihf IftdgUagp. 
Finallyn there hFis been n ftwd Jr^^l ri^fbnguhtic Inquiry kiiln the Innguage fsecn fi3r 
FjLEutLpJn^. |LAjTiS|, |H^?^|. ISim^.l], |NaEH6]l, which increftsr^ thr rhamrn^.T thr-it a 
parser b&^ed ais this theory will actually lirrfomi ^'^lU 



^ Jfi}i ti^i^ ' JrrliVEil 1h Wllt^'r A pjEJ-cr based <hi DrFuiiif Claiu^ ^rimmar iFinL rij««n thr r Alreinc 

ELruLK ]?fnnijLaLHm bund III fr^^-wiird tudtr EojifuapiH Tkr pnisei a wrULen in a ^eoernJ proof 
ij9l-ciii| null tbuT iiJlIm \f^V\ ^ ]ii^k nif txpUaAti^E h* pi^^ri , whjch Johnpoa does nrknoi^tJ^r A 
p^p<^ t}iAi prurLdr-a a \ru* HtjLino Id thu trableiTi inijai iiutpuL llhaUJtcu'dlr n>^t|vmEij stfkjbitkjrM. 

L^br]i4JhLflLr^| LTi b^c-. |i 1^ v^i\ ■LLiulnj to Lb^ LluGrrtir^ [nvu cf Lh? pua^r }>tr^^ETn| iiprr R^^^rtri-, 

ciJrtrhiJinU I'^KfUr* flic irvn^ rcdhiAflu on LfU. 



I[i'nM^ an]LLr4-li> E'«/uii|»Il- iVrhiEi Lin: Ui^i.^1. lAiiaUftrit'<>l'Mpf |MfM^r; rfJii^Lifcr f J )-^' 

fll ;Vi/r?/w(*- iJif ^-^i-rfYu-r!^ jFCrJi^ii-nii A- KjiJif-tij karif, 

J-EBG [UE-fici'-N-:tcl r^lir rfi>.^rjiir diifiL-jjAT l]4jui]it^fttij 
■1 iai\ lakms tte bofinir^maig Crnm the cTuJd-^ 

Thp fir^i wnrrf oE jl}. ^rijj^jjjJt--?Jfl, i* thp subject! tttf Eftnt. wcftcI. toj^Ji, ij fht 
i>l>jn-l: itiij rJir fnnTC^ Vh-uriJ. A-un/ir-frit, ]k rfin indirect ohjcct. Tib*' r^nfmrrffffirol 
fuurtAOD* (c.r^H inhj^^iT. ih^Jht^i iml imlErcc^ ubj^ct) i>r []|*'m' tvr>rd^ Fire fN'tprjLfcipL4>t| 
bj" iJirir raft-iLLArkiih^s [t-S-^ 'jJfl nf nr^^ffJiF-WzjJ, and aat by rlirir ptwIllciHT .-u inn 
3a7H EiLRL|->h- TIlIm J^ ^^ it^[i]|9]]|ird h> thi> hejitencj»j< Jn |;i] whirh arc tfjiiivalHCLt wnj* 
nf sflyii]!^ II). In rliHh*' ^Hjfcl*'iirr'< I lie nounN iiiuvt' nhour Tt^n^ly, Notice IbaL tbe vE^rht 

J?Hr?j4T-rJT^. FippPAIS Ul rflffHrE'lH piijltLUltH 4Ud VfcU^ ftlthouj^h flQt ijf FtlOtUt rAt-?iJ hpnii IL 

pti^j a|m hfpirt rhr ^rntcniTi!. 

^!^) a. Kuril tn-rUi^-riHj i>?fr\ti\-n\r jij^wji^ /u - r/u /'h^Wu-^tt- 

boaj]ieiAnf^ JMt-l-aj-l-.l iF^kc ^C».^PA.iT E-Lkc- iJiIJiI-Dj^t 
^II 1:^ Ihn.' booineiFuii^ I lUJi taking fn^m thr rbild/ 
h AnJ-d/jj-^'tf ^-ir- rn*T- r'li n^o/uJii-r/tt tjiWa ;? rr j^ f a ■ rn p- 

duld-LPjlF 3MJ-KHF I -J l-EHG boCiCI]eranB tALi>-r^4l^|'A^T 

^Ftciu r[iE rbilil L ajii rjikiiia (br boonnTFine.' 

r-EHl^ IMpkJll-[-;i rrrkr-PfOyPAST hoQIfcierflJIJ^ 'iLild-Dj^T 
^1 in] tfikLjk^ rite hooitHTfttig Frani rhs cbitdH^ 

TbETfii-i, howpvr^r, :kll or^jr-rici^ confitracat shoivn Ih rhrsf ri:4in]plef . T he fiu Tiliziry 
word, tff-rrjif-p'jT, ]ni]Rr nppi^ju- m tht»*^^iMi*L [Position.* Evetigivi^ ihp fijccd pofiitkjn 
ottht auxiliarT. ihet? four pprcLiaCntLi^^ rJo nyC ^idiEnisC the po^thiHrln^S hr uUcrbg 
rh]& sent&Dct- '| h*re Af*' bi FiNt-r ■|\ jjr ".i4, dilTerjjcLt wav's rj saving tht ^anw thlnRr 

So fftj wehftvd h«^ri tftl^if|fi ftla^uL niraDin^, and nor th^ p^^fs^r's f>ul|>ut domains 
lynt^j- Thn! cL-icm hrrf i« that tbe A^ppttg of itIttuiIti^ ihai r^i^iaiii fn^n^tAnt ^itiEs 
word paririiutni|i>A arc dirj:rl[v niiirnred in sj-ntux- TJifA^ tn, tln>cirdertnBCjrrh<- wtxth 
in mdrpsridftnl ftf thtjf grAmJllAhrAl runccion, which |i [atrr inlrrjarctcd eu part nr 
thc> [iLuanicft af the lent^nci*. In wdtr for a parser to proptrly bundle Jrei' wtird 
ortWt It tclusi rjuLpul the san^i? syntftcrtn: stfunrlUrc for ^Ach ol" thf pf-rtliP^iationH rj" 

^TJiptil KffhFlj, t*icrf h^u a[> wjitlen WuJptri \^^Vf ihr Warlpiri nr? Jitir^t l»*U|iht 4i Wrblca 
IJ^i^tii (liflt ttwiREwnan rfinrflflrrs- Thtp*" W ^Jhu ji ^tnndvTd ariJiEJ^rDpliv f« i|ir nhdntmri di*l 
da |LH>i opprot in J^ap^li^k r' tf . nj in j»[l to d«lClt^- j^ jKnfcfttalijrJ au^ luuiJaf Lci clw E^l|ill^h finfl. 
J3t* |pf Rfl^n I r^ ■ cinmptrl* dtHTjptJiwi-J Jhf Ufpfln-hM in cJi^ riamplM tr* uu pjui n^f lb? ^ Landard 
wr^trJI yj^?tciit| [krj juc ladiidrd i^]^ in ^ihI the Jio^irc Kad^i. 

laEbf nhnri Lur ]^ iiiJiLirilliilil^ firiin Lh^ arnlrbcr iCielfjuKL eiijae li-r itLT4D«l Tcgdi LDOttxt Fof tlMhPfcii^EV, 
drlijuir rrftitiic^ wj^] tir Ubrd, 

^ rhrrr u %aww djrJ^r^ibVf bmrm^ l^i? ^cil4chccv, ai Cfiviw2c, fytU iL mncwju a cbuiEC ib (i^cvri. 
tvlftrr Ehiui « dun^r jd i^w^JiLnf. Jlir hn4 wdttI lh fivi^ a lEifiri nMaE'biiD vvrr ihc otbcn Thh 
lUbLlr ilbJhnaiir b muiULli-i jn tSu EjLjIlnh tTTUkalaliJiJi* T^ir llir ^^ntni^^ 

^Ai'Lipiilhj^ iF bbV fluprnr In tbr lint fi^mkin Im ■ Tb? drEiilj ^t t,iLjirt rcMnptx. and tbt^ uf 
iJH.fhlird Lrluw 



la 



^ 



idiyk sentence. '1"^* n-^ult nf parsing 1 3} Antl Jia perjunt-rd couiini iJmtJd jield 
^triifturei tfirU cnrode Lh» firJLinr[i?itira] fu niitiuiig shi^vn in (,1), 

[S) EubiKt nfffljuJii^T) 

Thii JiLllity lo parj? lufh fHH-Util^l*-? hpii djJtKl pcrvioua panlng systems- tlldr 
djfGtuJty wKh Frrr Word order cnn Jir i]r-n.iLJiur rated ^hJth a ¥ery i^i^Fmncn pnrBlng 
tfi^lhlLcjIngy. runtejrt-Ertf |]Ariii]BT tbAt ftruri^ fnjnj cjuLpLlff design." Ctmlrit-freEi 
pFLJ■^b'^h JMT h^Hid i]n rtbritt^st-frM- gr.-lEhUl]A^^. nin^urins irf :* wt Df rewrlir niles. 
The^fi j-Lik^ nintEuii h1 letl-hand mi|i^ and a rifiht-lii^nd flide. On thr fr-ft is a sitiglc 
non-ternunai nynihol wbifJi may la^ rejd.ictd WLLb tlir f-iwin^ of iyrtbFKil» (teriiunal 
frtLd tiocL-lernLln^LlJ or\ thr riftlit f^ivn^n A spt-ccEd Hfjg-t syaiilml. Lhe ^AtEiJi]Eu- is seud 
trfc Jrrdvt a string |[ Llit^rr is son]e ^4jqi|fihrr J]f rcwritf-i lI^aL rc^ulti in a sH^4upnce oJ" 
termhtiJ ^yn]l»als Lhat Ji^ati'lit^ Lbn? ^Itliir- The L-uiguFL^e of luffi .k Erninniiij- ]s the 
set ofstr-hn^ lliEit il can deri^^ through all pon^ihlr jcqa^nces of ttriv r]tf^ . 

Conlcxt-Eret par^rr^ suffer froin Iwo |irrihlcniB when it coi3]eB Lo parsing free- 
word -^rder ranE^afiei.^ lii>tEi resiiJt Irom rl^b- [i^mre Qf thflf underlying firMiiniar 
rormaLam. thc^ rlrsi problem [9 that i^xLremeLy iqip pfs |i i r qfjice ^j-unjii^rs n]U5t be 
■ATiLteic in order to covt-j v^i^rnJ percnutat 1 ufl TJie*e ^aa^unftr^ bLJe the r«gnJEU-ity 
behmd grftttucLiickal Eunttioji^. ConsLdur iht RTftniitLar i^ H) th^t rov*>ra a lanf^u^« 
rrmLaining er^^lusivdv IrBntiUve v^rhs (r-^., s^^nt^ntti K^s\si^t of a Tfiril, ih Huhject+ 
ftJid 4in object) W-Vli IhL^ grRimimj- bii mlf? iiTL^ requtf+'d tu uhscure]}' n^cLccHdi: the 
fact Ljiftt vorba take two ntiilhhi nneof wblebu IbciiibjecL Etnd r>nc of which U tinn 
object. 

<4| S -. NF, NT(. V 
S - SP, V NP., 
S ^ VN?, WP„ 
S -^ NP. NP, V 
S -^ hP.r V NP. 
S - V NP. NP, 

Thrsc pajT^r^ liftv^^ i cnore signi fi i^af 1 1 fsiilinE- Thn^ ^tmct^irea that tb.*ft mjtpnt 
arc not Irngtl i^M r,tlly pieciiP bn^cftas^ ibey d^:i t>L>t n]?d:c eiplitit lmpr>rLant ^yr^tartic 
rnlations. Th^ «unplfl grarojiiaf dor* not, for LnslMq^^:^ hishlifibt tJir ymnunatiizal 
functions of subject anil o^>J^■cl. A better ar^cimftr would tnrud^? ttiiB infunJi^C icm 
directU qgtb ri? the hieran^hira] granknLar fiEvtn m (Ef. Here thr subject Ji tJi* 
whljna of the ^Mrb phr,isr HenoEed ^VP') and cbe ohjett h the sibling of Lhr vwh. 

(&| S - \P, VP 

VP - V HP, 



TlT^n „rv If *rr«[ <]thi-r qKilital Inap^uo^ pAHrr» m Lhr Eri^iat urd- wlikfl rtllbDunh tbrj Jiir nuL 

■crriufjly l.rtKd DTI canlf^t-FJnr r-iriri, iti]l Inci tlir- -i^iiiri |{] tandlc frfl-r ^wd ocdcw \ lirirUH 
hb«Ld lindl> !■ thr LunrCLulin^ ifulitrr 

11 



Hoi^^vrrn tlil^ BrAinillUT hiilfH-n frcmt LRF^ilrqiiBir rrTTprnp^r. Etud mnovini; tht 
ordering Qf the eJejuent^ of iho riRhr-brUMl iJUfT ttiL^ iStiinlrrMiJ- ^^]^^ nnL f^PQerjili' 
richer oE ttir^ ^cjblenc? HclieiiLaia fouDd In (ti^ 

(tt) VNP, NP, 
NP^ NP, V 

My $r>li]titiii[ (syr llic pnnblirn] rifpFir'^iii^ frc? word order i& ha^edart Gc^Vf^rtllil^tit- 
Bmding (C"JB) t]iKpry.=* t;|i i-^ a Eid^^uhlu ihrtnrv tJi^i i:^ f^finrcrnrd wilh iLe hTntar 

of A Eiiigle sent^n^e. The "iCructLires thiLt it provt^Lt^:; wllL IF it In tFHTT^i. ninkE 
|]|e 5(U|iort:uLL tin^ui^lit InrH-^pliailOEl i^;c|>liH:]1. Hoi^j^Vct. (IB Is not a coccbputAtionflJ 
LbcDry. ll does noi specify Ad^u' parbini^ (or ^eni^ratLi^rih, Fur ^h^t nu^ltrl ii lo be 
donei ]t cholv "^iM^rifie^ u^€t tbe underlyuig aj'nla^ Lb to Ij^'. Hy b^^tlhg ihr parser 
on (;B. I mi^^i thar tt^ iJ<itpLH is dkrrriH^ l^y CB Ihi^jr^Vn Eind PtirthercuLirtiH tltAt ifif 
opfTAtion-^ ot ih^ pAr5.#r Toljow iS^ morfuli^rity oF thr linguistic ib^Qry; tLis mU be 
rl^hcnrEitod belo^y. 

III r?icL cti? I^rsrr cnill^Uit4^ oelIv ei |??irt of GB oitt^ut r^pn^enlaFion^. (Ht E'^hn- 
sift9 oE aevprnl Ici'^Il oJ" j«p r&hon LaLil^k eAi'la i>f wliLdti eji^ihd*'^ I ti FrHriiia I Ion r^'levEint to 
a certBjn A^pi'ct of ihr^ ientence^ Th» pajher prf^dn^E-^ Lwo niuLpiU KtrtLdti]J~^f hn^ed 
on Iht^u kvcls, Prflccikji^*' slnirliire (PSJ repT?seiU& a part of so-Lalled PIhjJio- 
logcal FrHfLii 1PF)i a« ivcU [i^ I^H'^rl fx[ Lbe Lh^;i>rr of roorphnliigy ^ word struilure^J^ 
While syriCacEk =:trm:tiiTT |SSJ reprt^ent-^ -irt-i'^JJetl S-StmrtLirc. Thr HirnrV hcbiind 
tlie±e structures Is qiut? coniplex: they wjJi he de^criheil in ^r^Atfr dt^lhtl m I hi' next 
cflRpl^r, Fur our |>ur|]05n hi'rr^ hr^hiFiTrT we rftji gi*p au Abhrevl^led di^^rni^rion 
LbftL H'HI s&rve to i^lkfhw Jlow th^ linRiJU^tic Ih^Dry |ircffirt^S Ifir Jintp^^r s^niclures fiir 
hajidlini^ free word order. 

Prr^tdrcLcc ^tnitliirc is dsed to r^Enre^vnt lb? prccfldtnce information inhef^fit 
IP the input, whi^h I Lake tu b< » silstitly professed vft^icKL of Ihr 5|3«ch ^treaiii 
ll i« Buiun^d thai ihe speech str^an] h^\s h«n broken down Into \ls C4FI tl ] t Lent 
nir>rphmirsi ^orcUn F^nd phrFurs, upon which iherr Ii a ^otal orderinj; by ^'iTtu^ c^f 
Cbe lineAr nature of speech- PH rf-preient^ thf rtrdtTiJIg i>rthi^ IrLpiil cLieiL iH rclovant 
to syntax: thus ii is a pArtial ordering. For exAiriple. in the i*$ For ^i^rLjfaJu- rfu there 
Wuuld ha nn ■^rd'-T^rj gn^r of morpbcmts ( rt^^Jn^ir, rJuj^ as the ordercng of the nnun 
foJloW^I h>y llhH^ c4^t'-i[L4rkrr h syiil^ictkALly rrlrvHint. In fact^ It is ungrnmmi^tlcA] ta 
revr^ri* the order of ihfjJUirplltEiieii rli^ri^Js tir^ »adl Wr>r<l evs *f (h- ri^/[r/u in Warlpirin 
On the Qth^r hruid, the PS for ^Wlpiri would not ^ChUtAin relAtioru: hetwf^tl thd 
WiJidKy brcausr Ihuir i?riJpT docs not niatter.^ 

The other part c^f the pArser's output i& ^yritAcCLr ^truttnT^n I'nlik^ PSi SS has 
no prcredcnc? InfornjEirion rncod^ into ]t: it i^i a strictly hjeraK'hi^^ ^Enamrcr 
IhjL diifefi frflni tradittojia) <:ti theoryn whrt^ S-^ttiictUrp h nn ordered k7el of 
r^|>ri'¥mtiNliitn. TKr arK^Jcnrnt f^r rrnao^bg thj» uiformatLon rehi^R un Ortajj]'^ r-a^^t^ 
theiE^ doe^n 't s^us to be a need for pr^e deuce jnf4^rn^lLLhL>n at th^l^^rl uf S^Hlmcture 



^^HTT, fw rwuniiL*. |0liu5]|, ^ Ini53. 
ThJn u flbiii 4 hjt iiira luntitiiUfl^ idei Word QrJ^juip; ifjcSilh pfaroMifr dan nutlrr; h}i9^ h handled 
\if ib^ p«nn. uH vLiI tr iln^tdKriJ in Lhr fi^Jij^Lnp; diipE^n. iL u tebUt LIi« phrun i[iLt nihF 
]»rJiWt^ Pr^lhhpvlt ^iHltMtT nut? Urqtly H<-^lir>lr [■? r«U Warlpm a Eree- phr^mnrder luif[urL|c. 

13 



w^ ™i c?T6f 3S iiJiipLj- fL5 hLwnrrhicnL SS ij Y'bai^ icjcli rcf^tiDiif Bb RrAmninLirFLl 
(^mctioD^ would lii^ trpr^-^ccLtrd. Subjects h:^ inilrii io be bhe BJhlkLHi of VPe^ anrf 
i^bjrcls iHve takeci to li^^ ^Llilin^ of V. 

Let^^ ir^i h^w th? CB ^aniuiAr wuuld n-^ccoiint Ibr tht smL|>lr JanBu-iRe fif tr^miJ- 
tite verba Jn Wat-fpiri, GranintFLticjU Fmirtlnti h idtnHiied bj tKn^ co^r-mcnti^tir^ [jtl 
iLCPun^- Fni CHIP cla^fi ol" verh->. Hi*- ^nhjcct ia itiiirkn] with ch? ergati*? t^Tr cnften 
fipp^jirltig ^ chp juElii -r^ii5 tbp <ph>rr 15 ]TLdrked witli [Ji« ftf>^y|ii(ive caw^ ^iiirh ib 
not overt plLCjnokifik',-xl[l'.^" At has nJieady hr*^ iiirnCioned, i7j^ar^ ciaarkHzrs must hi^ 
filclitk ((.e.,affiKHlMo tht lirmna that they i^wl;, ncid they miiit b« to tiieir nght- 
The ni^Uh^ so marked receEva L^e fa^H^ df ibeir caBE-j^Artl^r. V^>rbB must h*- mrifcled 
Edf ten*e }jy cr^Q&ff majkeri fbftt Jut HmflitLi: to the itettj aihd ten their rigJit- ThH?*e 
FaJiti are Encoded lu Ihf PS for Warlpuu ^ slluWn in 17 J. 

(T) PS" N ts Followrtthv C {intraworrlf 
V 13 JiilJjjweiJ Ijt T {Liiiraword) 

Th*- srFimii]jirn:ft] fun-^tLrtn* of rbv nouns ar* n-prr^rnted Lq SSh As wJth tha 
lauprovrrf grnccknwr al»o>F-t-. vi\^ i'tiii-. [tier a auntente to frtHniKL of n dgliq plirrut. ibe 
subJMtj AEida verli phritsr Thn? Vl^ ui turns umsE^ltol the verh wul a n^iuiiphrjiseT 
Us hlJJL-cl. The S^ [or th^ ^iiMjilr Waripiri senti-tJ^'ra b J(ivcn in (B}h 

IS) SS; S dunuDiLt^ NP^ ,-;nd VF" 
VP Jij[[uiiace<i V mA yp^ 

KoW £oiiE|d?r \9} *hjch is an abbrei^JtitHtd vrrKlon of the iflflip^ araiteacss ([} 

'I ai^] taLing Lhc hoociiprAng/ 

Tha P_S returnnl bT the paisar entiiilr^ Ibr chree orderinfls frlr^nnl Id synLw, 
iiM nf u bkh Ar4j Lrtrjiword- ^fr ligure l.i. In this jsrJiptdr depictboo o| ih*- PS, ncdas 
leprp^MthtitiH Ciitc^rEM are CDfin^ftrd hVhr>rc mjerini^ is rtfcmnt. Henca th^ Lrpc 
rnnneti-ing the vn^fh hic]]]! ^lq^q. a^d the tcnsp elaniiints rfit^ [h^T Luslanca. ficyXr 
Ibril ihc urderinR b»*lHH^rci rhc word^ if ucr\ rfpc^iM^lf^ in the Pfi fof tht aeatence 
hanause It j^ unt iniporLanr_ 

The HS tlittl i* rviQmed rojitiijtll- thn bicrarchical sLructiari- Trnr lh*> TSfb and h&th 
it* subject and ohject. ^^ h^iiP^ l.a. RenmEcib^ thiit tfn? Eraph heie dc^rs nnt use 
precedence ai^it caulcL ha^? been dcpjrtcd *i-'blh ihE-^LihjecL n^jd« and the verh pbrAV 
Flunlr in tbr' fjther urder, likrwiac for the vefh h^Ar and the object node, 

Uk Jnruufl^Fi lilif EDiJtdl. Fr^r Lhf putjjmj itf it-u t||r^. ih^n- u do iL^nilifAnc iMcit[i« nitqcT 
tlun ifl ih« JabrJia^ nf tfvr ■ rwF-iLiirki"ji_ 

In FBrt, hlu^ srDEWiir dk tjnKrniiLniHlml tKniuf tkv rwU rtUxLllax^ wrwd dcirt iiot Hgr?^ wHFe 
-^Ifjci-I Elf ldii- viiL^ihr ^TlltF iLh FJiib fil ^hr ADidicLrr ^1 ^1^ gai^D h^tcw 

13 





® ® ® © w ^^ 



FifUTE IJi Tk€ P» (iAt ^9)- 




^riEa-mi 



Rgure K^: TN -^^ for f9j. 



1* 



EnfiCi'ih, thi>rr mt ni3 E^V^ft ^fUf-iaiArkLTi Ebr ttn' subject or th? objc^tn In^trAJr thr 
Eiibjecl prerrilrs tlit^ ^ffh jmd the object foUowF, In ihls s«1$e^ ill« v^rb Itstir atli 
JL!< tliH^ ch-^e-niiirlcer. The ncun |]hra*+' i^ (ti kfl i> iTiart^d Icr DochunaliTC: cast, and 
tKe noiiJi pll^&*^' to it^ rlgfil is [TUrfkfd l-^F aci: li-^hl ] pe cam. As wifh WarlpLri. Ihc 
Verb ih in Heft ^1 fEhF rect^e with a tense ekcEipnl that ]e cnrLjtir Lfh LJir vfrh i^tfjit Juid 
to Jtfl nahl, T^tHw Edicrq are givi^ji in (ID] ^hhU pH th? Ph for thia iUiLpLe suhsel of 

(lU) PSt N._^ L» fpllowedhj V(hiterwL.rd} 
Vair i^ ^iJ^i^edeJ hy V (iDL^rHord) 
V iy ruIJowj^d hy T (intriNWitrd) 

As *l1li WiuJpir]. lhp erFLicimF^iLrn-J funrtirHti^orrhf: n<MJP phrasBSJire represented 
m RS. In faci, <hr liJa fnr Eaiplish is thei^tiaaa tliAt for WarLpirL. The only ^iLffrrprt"! 
concerns rhi' niA|^pi}?g Troith CAEe Co gr^niohatk^l funclLi^n. In EcLgli^hr ^h^ tloii[L 
phr?™^ nmrLed Tor noiaunfltivt Jinsc i-F niftppf^l lt> Auhjeet^ Jind t<ie nnun phrase 
niAjked It-r accusiiUte lUmt 1$ iiiapped to ohject. The SS For thr ^[[itptr Erijlish JJ 
Bi^en in {I I J. 

jll) ^^r "^^lonunfttei NP, and VP 
VI' *Lo]idnatc* V Ajgd M^. 

Thf rr>Tj(h tfAn^lfLtion ofjH) is given in [Vi\r Lct^s ^LAilun^ the iwo structures 
returacdhy rh^ [W4f ^i-^ |o r t lii* stn tencc . Thrlii^tn PS^ rhH^c^dt-s the oidefinR pelAlLons 
th^t tu^ iynliwrLcally ruLcvnni; brr figurr 1-3- A-i -staled ip the PS foF tht5 suhact 
o£ t^^L^h. the vcrh 5tem ailH^t U^ foHiaiMd h^ rk tense element, in this case -[ny. 
rcTi now WL' h^ill igni^r^ ^he lilocIaI verb am as it h parr of i]i4^ ^UKllt^l'n which U 
n4t belnfi covered by [he !Lin|jLr gracnichEvr. Th*^ other pajL of tbe PS concerns the 
sijhjerC and ohject noun pKraicn, Jj^tirJ i'l' fcr£p?rjt™n^ respectively Note thiit in 
PS the ■^trln. fr^ibra^ is conuected to the noun pKra^cf a« Ihn^ir r^JuMve t^rderipR J^ 
rtltV^inl, 

(IS) [ am taking thr boc^En^rAtl^ 

The SS that is returntu] irunl^ins the hierarchical sLructurc For the icrhand IjOth 
its subj^-L Miril E^biect See Eigure lA, i^ is edjuiyait-nl to tll« SS EoT the WarlplrL 
**Titfnce. as one wouLdeippct. BoIIl sccLtrilCH-s ^v^ ^a>infi rhe sante thbiE, al least afl 
far as C hi: rough cran^latmn ^nnj^a. Inasmuch as the meanings i^i^ rquiirAlfpt we ^>iu|d 
cip^ct la ii^e identical Bjntarl]c ^t^u^nJ^^^^^ wfiuHi the paj-i-er does jndeed provide. 

The dlAerence hetween the so-c^LlnS fn?i-LKifd order language. WarLpLri, and the 
fixed^T^onf priler lanfiuagen Englishi then, is whether ornr>t l\\^ pfedicbtur {Eor these 
Klu^ple sentences it iF thn verb) i$ iJ»* a rise- rctj ker. Tn both langurifie^— and. it 
in bnlir-vnl, iti 4UI LnjaguAj^e"^ — case-majkin^ is a dirc^lrd rt^lAfi^Jtt- Iri W'^rlplri^ this 







V} IT 



FifiurE Lir Tin- PS for fl^)^ 




fUltfRf 



{Ap A'li'rTicnLEijI 



Figure 1.4: T^e SS foE (12), 



IS 



t^fo HiJ ca£f - m Ftrki ng 1s ^hf^rrorn^cd by cltticB, ojnd tn En^li^h by wordE. Wbea tke 
pnidkaltir is Iht LMtJ-iFiArfci^r, llitn^ 1^ *n (irdw-mj Hlttlon tk^C^e^n til* predinAtor 
und iti^iirgqment^^ when Lt is not, there is noEucb ordeno^ r4>lzX]ciii. Thc^ ditlcncli^Mi 
h directly leHecteiJ m IIle 1^*0 outpat &truc Lu rE^ii. It Iti OiiLy ^ queBtJOi] o( wbjcll 
deniGiitE in PS p<irEcirni IIjO ordered ^tn>u oi taM^iTiafklfig- Tlie &yntajrtic relation 
of firtilUdaiiJial [ilJLCtloJk— r^prftsented in SS— is unordeied in bath cases 

The terms "^rree ordef" and ^'Sxed order" are a bil rtiitloidingT lLi>^'CVHirr Afl 
noLed abo^e, WarlpLfi dws EKbibil kuire: ordEiHug plLCilc^tncn^n f<^ In^thnce, ^mong 
thf niorph^m^? i^f a word, and arnong tbe words oi a phonoloj^cal phra^?. At 
the fiame time^ Enghsh ex]i]hits soinii iTeEi order. Oile: 4.uirk]3iOfL d^^acnplt JS the 
ordanng of pn^Ceitional phr?W^J- Con&ld&T the &entencea Ln (13) and (L4) which 
mean esaentiaHy th^ &ame thing. 

fl3) ] went to t!L4 fttt^ti wIIIl Mlt>-. 

(L4) 1 "A^nt with Mary to the Etore. 

PiepoGLtbonal phrases can be procJSBed -quLtc nn^illy hy ttiC p&ff^Fn TJte cnr^^red 
rdallnjn bci^'rtiLL tliC pttpMJtkfcfl find Itg object roup phrase ]& given in PS, Eho^^n 
in {l^}. The prepoEitEon is a case-mar^?rs marking its object fOT lt& own Oi^^ 

{lb) PS: P \S followwl h^ NP (Nlterword; 

Synta^rtii^a] I y spehkLngi pTepoEicional phrosct EnncLion a£ obJE^ttS of the mair 
v€rbi EOj lihe object noun phTasc>«, thi^y ftTt doiulnated by th? VP node i^ SS^ 

(]6) SS: V? domlnatefl PP 

The FSs Eor th^ si^mple senLences aie _E^ven in flgnics 1.5 and 1.€n Notfr that 
the reloiriL ordering o\ lie praposHiOnS lo t^e objen^t noian phra&es it, indeed, 
represcTiLed here, while the ordeiing of tb« Verb and the pttpOftJ ticitia] pbra^^ ii 
not. The SS for both oflhc seut^^t^Cti bl lihcvn 111 ligure 1-7- 






Flgure 1,5: The PS for [UJ. 

TIlce njinLdn] of fsxird and fr« H>rder se^FlM tn liold acrces lanRHagK, Uotanguaje 
GKhiblU *lltiwlj Iret or lixed order; rathec, laaguaflea Lip aLong a spe^^truin where 

11 





m€mi 



wilK 



U-T 




He Afun: 



FiRurp L6. TbtPS for (|.)). 




birnf ^1* live Hare ^Blth Marji 



Uglily !.?: Thf Sl^ for (1^) and (L4). 



» 



WriflijlrMJtH al ojii' EDiL -lit J I'^T^Bh^h at cIlc othc-r. This tpndt support Eur 4 hiparlite 
TEpreMDT£Ytlon thai permiLa proco4ti]Clg ot bclb. The dilFcJPJiri- in prtK^h]nS p^fCkc^ 
ukr langiiafiJaj ^UL ln_^ rfJlected ir whicli rlTrnciiire do-^B wtikh pat"! of the parsing 
job- In more H^*-r&] kjiRuas^, nii>T^ i>f the ortLerLiii^ iai PS will be relevsiil Uriyrt- 
tA^' ID IcBS ortltireJ litL^isAgffi, lei£ of thp OrJtrlUft will he ii¥i]^DrUnt. Ifl hotli caaes, 
I ha K^ntACtlf -^tructiire wjU bear rVc rrtponsibiljLy Tor K^srH^^nting t lie grinnna-tkal 
function oClhni H^JeinentaT aa well ^ uthiif synt^tLH relationh. 

PrOCOS-^ing ElPP- and fLxuJ-OftlcJ- plienoinena ar^ ilOt tht only accoraplislimutlla of 
Lhepaiior- ■[lieconLribiitLi>naolthe thesiBarflUslH^ briefly io table I.I. The pii^er 
Jt>ips a new ajuJ grcpwln£ class of ntt U fil- iangua^ padiL^j* hftsed on C DvemmCiLt- 
BindingLhmryJ^ A ma.jorlaak<>fllil^ ttn^la waatoprcnllsi^lvfi^rmulateGB tln»ry In 
Dnier 10 bniihl^ to compute LlE repr^apntAtimt. In thw process^ I foUcltl It neresEuT 
to modify th^ theorj ol S-itrUJTtnrf^ (e-tplii]tcd betow} to accaui]t lor fiee won] 
oHi^r- Given lb]E actuunt, the next task ^M to formulate a tel of represen tatitHJIS 
and algoritltJnft ihal would tompi>te tbe Fnappiu^ from Input Beutoilte to output 
itnlCturea- Lastly, 1 implf^mented thia do-JgilT ajid leated it lhjI a range of mputfr, 
botk erftmniatkal ind unfimramaLmal^ bo enSuW iliat U handfed the phenomena 
pToperlyn 



baaed tb^ parser on GoVJZfiin]ejit-Hknd3nfl tbeorj 

modLfijKl the tlnjory of S-&lructuri3 

d^igfird t|i£ n^prefienlalion^ ^d ^n^^^^J^^^ 

ff implemtlittrt aqd tested the pir*er 

Table l_l' Tbni contfibuLious o[ tin? Utefll&r 

Before Introducing Ihp^ pa.r&er fonu^Lyn iti thf rollowLn^ Bettic>n 1 auLHne tbfl 
Warlpiri pltenoJIlCJka tJiat Llie pansH?r LAeI hwirile. 

1.1 A Warlpiri Primer 

WhlLni tliere aie over 2bW} native speakers of WarlpJri, Tew of thcart will read thai 
thstfr. Thortfor^^ It la useful to iutrwlu^soine twbk Warlpiri- If you ir^i fluent Ln 
WarLpirL, fcel free to Ehlp to the next aectlort. 

1.1.1 Th^ Major Parts of Speech 

TllC primer bepus with a dl*CU5*lon ol Lhe thrw major pitrti of ape^ih: uounS, 
verbs. iudauxJllariea- The acialjFis<pf tioun-^ and verbs ift jelativelv striUghtforwardt 
IjUt tlie auxiliary is morr H,Orti p|]<:ated _ I hegiu with nount. 

Nf>ijns mu&l be declined, wUh the unte-marker sufli:(ing onto the noun. Then! 
are three tiyntictk ^W*^ in Warlpin^ 4rCg-%livp, ahaoiuLivt, and daiivc. (171 filvefi 



^^TfcfltU irt 4 ni^HiHer Drpar«Ti boK^I Utt CH-llflr hnaaL^Llr IflfiOrlW. Tlltre arc dh> * fa^lifDl ni 

Ciil-t-AHj^ pnr^Cr-^ 1 il|Arii>N Llin^^olkrr i irnwlrt in 1^^ r[ipdiiarlir>f; 1 kbpUir 

19 



40me esamplei of dH^iltienl nounfl. (a) ahowfl ^ ntjuft inarlzed For dativo tMO, (h) 
ahowR ^ pnjilOHrLj whn:h lb dedtitlcd juflt M nouna are. (c) siiO-^S a roUTl marked for 
ibpftluttve case; Han^ t^ HO m-erL marker ofthj* TMt 

(17) a. i:urdii-lu 

*tliild' tnarh«l Tor d^]\f' catift 

I-£KG 

'r marled ror^rgiLtJvei^Me 
c. kurli 

bcGmeraog 

^iMXHrnWiLPg' marked for il5a<J n ti ve taw 

Noun* m*y b* declined lot" ftumW, in addiLLon lo cMhv T h^re are four aiimb*r- 
marlif^ra in WarlpirL: SUlgnlar, dual. pAutSil/'* ftftd [hlural. Both dual Aahd piYUC^l 
tavfl oVtrt phcmological d^mStilfl w/tiKh appear jual lElLf tbe ncmn^ enclitic to il. 
(le) J& An eKajnple ol ^ nOtm that If overtly marh^ lot nilinh«+ 

chtld-DITAL-Di^T 
'to/lrom the twd cliil*JFr[l' 

Verbs iiiuflt be JnHecled for l^ru^e *jil1 m^^- Thi& Ls Miiflnipllslied hy fliiflixitis 
a tense elemcJit (wlikll contains bolh teoK and jnt>>d ijiformation) onlo the verlj, 
TL^eftr^flvetiDfieaii] WarJlaJrl: non-piRt^ patl, arr^aiJis, futurt^ and prffientat]onil. 
There are alfic fi vz cOit j^LSfttlfin cUiseain Warlpirip ?0 th^re are a totil of 1^ iniecUoiifl 
lor vcrbs,^^ (19) prBBcntB a touplj; ol esiunpIeE o( billected verba. 

{19] a. punld-T^t 

'is taking' 
b. njTfl-tijtJ 

^aafr' 

The auxili^y word \b aiaaloflous lo l\\.^ aij:{|]lAry verb Eyslein Id EnftlTah- Wail- 
plrl^i auxiliary consiBtfl of Kcvi^ral tonipDnent morphemwi, hS^'Ii of which ie ciptiH?niU>' 
ulteiEHir Th^ ftrfit comporn^nl it a cOcnplemeJitizer wblth i« Jiull {i e^ not uttered) 
for the rimple dGclftJ^tl vt aentencefl that aro tliR domain of the parBer. Tho SKOnd 
cornpon^nl nzont^hii:? ajipect infoiiriitiOTi that comblnea wiih the leimr atid mo^d on 



^*Tte paoc^ nambcfik^ hp for rf^rinfctr pJuToJ ^Etrcni.c, h^h^tt tilt tH^rmtn aic prcKaL dlitilif 
Lbr atCrraDce afthi^ tl£p(4tK^ Tl ^(lald be [u=d ui the tTaii^hlkOII fif Tlnut ktet*! ercel |pnk]t||i|^] 
u-^ hTfiiLtliag boonl^tAnf^a " Thv |i|ur^ munbrriji^ e» br Uiit^llnitC [ilurM rrfcrencc TLju H^dd 1h 

^^Stn(N«flS] ECW ^H^DpiiJEtc labJr dC U[u« €kui«ILtBi 



the- verb. Tbt ]^' coin|Joiieiit contAjna r-i^mlnjil *grct]1]ei]t mformilioii that nom- 
bij.Qfl with t]he sul'iect [Niid obJH^tL ijf LIld verb- [^j ahowa &i| aiivjllary JjldJtilltig 
thrit tht ^p^?l:: i& imperkmL, tb^t 1-lie -^ntjecl ]^ llr&t- i^ Tucni sLngulart and thut the 
obfCCt IS I til H- |3cr-nC»n EL]L|^aLar Lr tbc dut'^E cjuift,'^ TfiJfl ftirju uf tLo impeileciiveT 
^H ii]&y b€ ii£€4] ciiiJy wklla ihe Tioci-pa^t t-n^jie. 

120) fco-rtw-i^Ji 

TH?EJtF-la-:ld 

JmperfjXtiVJ? aspect with Jiral-peraon &i n £ular * U hjct t and dative objecL 

When an ^km^nt <.^f Uic Jriuuliary ie no! ittLered lL does iiot vajilali rttUd the 
iJtfitence. InBUsad, there i$ & Hf^fault t^n^llihi^ for each of the rmn poneQ Er As 
m^tlooeEl Ettnovi^. i\i^ dri^E^uU ^"alue lor the CGinpJernenUi^r pnitiltlon Is btdt. Tliifi 
conLraEtE with other, ovtft HlOrtipl-^cntd tiSerB th.it can appear, fiuch ^ llbe negative 
Ikal bctb Jicope over ihe entire tentflDin* Wh^Xi tlit baBni is dOt utteredi the anxiljary 
Is given a poifi^tivc ^pecln wbicb Ecmblns witb th^ lert^e on ttit^ verb^ ad with 
ov^^rl baaeE. The romiiial Agrf^mi?hic flititu cLtf-mll lo ihi rd- per Bon, aingutar_ The 
imllauKjhary. pvt-n ji]{2l)^ if interpreted as cQiitalnlrRp^rftttlveaip^t icid Ihird- 
perBon^ Bingiilaj- agreement, 

(21) * 

perfecttve aspect ^^illi Ihj nl -penicm^ Bingular aubject and object 

1.1 r2 ClL£^ Fhra&es 

Belbre diacuBsing case phj^ajn^^ a i>il more WarLpiri phonology nius-t be llltr«lui:ed. 
As laT as nonc^rtl^ Lhr pa^Bcr tbere are hui Iev4-1& c^E pti i^l^^ilOf^loL ^p~ouplng preseat 
wLthin a Bentente. Tbe Loweat |av?| consist* or morjibemes. the indivisible nnita 
Ci the inpntJ^ The next Uvd ^-Olitailkii ^vordB which are sequercEfi of jnorphemeSr 
Not demons tratCH-l t^pliccLly ia Uie e>[ajiiplefi qn far Is ihe thiid l^veK ptionoloj^iiAl 
phiaacH. Theae phrafiea are sequertCn^ ol Words and are ideatilied in a Etraigbifur- 
wMil itiirtntr frith word atreaq inform^ttiort-^^ l]i tbn: t'xacnpht givea b this thsflit, 
anIeaB noted fitberwtsOi esch *Ord cOrrehpond& to a Hlngle phonH>lop^^*^l phr*4^. The 
Iftflt Injvel ia that Qf Ben tenrsfl ; the parser ii^rkii ^m onl>' one or these at a time. 

Like Kngil^h^ noun pIlt^S^jb hLH] not- limited to a single noun In W^rlplri, Stiuh 
pClraaea — actuaLly caiLed casf^ p^m^r^-*:^ tliCj" atu idenl.ihcd b> their caae-markiiiR 
ma-y f-OlhSJ^t otstiV^raJ nouTLB within a single p hc^nologK^I phl^e-^^ 

Tfar Jicmhnhl ^Fi^o^ni^rir cJJilcB arc lahclcd in lhtd ]:4TCi The fi»i pv^ p^FWi AT141 nnntbn 
iofDnnaUDa Tor ci^Eiip]^^ 'M ^iLnnlt Eor ill^t-t^crwri UU^Iv Ttie KCE^ad pmil u 'v' for labjecti, 
kod ^d' lar i>bj«iti [^lLltc ohjcrbc ^r^ ni4tk<Hj wjlh 'd.' 

^^McTpiierna n rr nDb p^Tt cif pliaiiciU:^^ nrdr^r. r^t}iet oC DlACplititD^ Tbc banc aactn of pIidbdL- 

p}lf>n^m^Nnl« inot^ibnli^tjIcnU^ LruiLvicjUc ujiil^ 

In IVLrCfiln, t}\^ I^Ft-nkn^i i^rA ut a phnuc tcccitci pninar^ ilKzu ibd the ECEiiAhEiJef rc^^f^i 

#MIWlJ*H titreftP TboB. It u A winple oprriLiim la<fM||piML t|l* t'Mt^nc of ^illtia^b L^flrJ Oh tbs btttiA 

[■liifm^tioii j^t?7C]iL in the Furlarc otriDg 

Thnc hit ^\h<f dbmm Linnftun cut plirue* w[iMrr lEiirrfrriL fi^fu ^ the r^i^ [sht^-hO £tL^t^^4t 111 
phi^notD^LCLJ phru« h^x^^^L^I \\y bt1kT t^hrHcn. Tluti j^cDimieDoi] m ewL ^cL iiuidjcd by tb^ 



Tfi4> i^rJL'rillj; <J ni>UriS w^Uiili ^ j^llOliChlogi^d j?Jii^r l^ crpti^lniiiii^l^ cnoi^Lty dui^ 
lOH:^<i m&rking fiinsidpnii|oHB_ Qch^n^l^e prLin:irli4^ nonos may heniarktd for cm& 
b^' ca^e-markcd Doanh lo Ihcir ligbt wilhoi] lbes[Ln]e ptr3£e~ Lhe case-m&rker m thin 
r^iAc II^ ati ^>rit]tdftl .^fc^pi; ovfj f^ll of the noujis Ln l\\^ pliKUK^, iml nil Jnsl over 
^he Doun to which it is ?ncLilir. \/£2) giv^]^ &ti ^xaonpleof Eliu phGnoniQaoii. AFarfu ia 
ivcyl m:irk^ ^r abanLulivf 4'iLR4i bFfaijRf^ it apptarn in rE|B phrase lilrmg v^UIl a i^^ifj>' 
■[larktd m>UEi. P,t., nnrt-h"r [I Tit'c^uM hi? uji^rftmrnaMi^Etl for iht fitt^-uiarhE-d nouii to 
Appoar tc the Left o^ the unmarked hquil; ihls is shawD Ln L^)- 

(32) FTurfu in'ri-ip 

kangaKK? big- Hat 
^to/from the bi^ kad^iirijo' 

bl£-DAT baiLgarOu 
The ^ynta^c of cor^ tLnUoliK casn ]^1trj^es \h actnallv a bit more ccmpljcalfHr thi^TI 

BO Inngas th^ laltcr noi]!i9 u.Te ma^l^(Kt by a CH^e-inaTLcH ni^mn to Lheir ri^litr^^ [24] 

narfu- Tlie Rcrond ditit'e riAe- rjiAi I^pt , -fa\^' Isas scopp O^pf Lh? gpi^nrf tww word*, 
f^tftup^^pi^ atld u'lTi- Tb^ accobd word, yaih^r/^, Ld ^owod El^ji^ b^au^G ther€ jk a 
[:££« iQarker La lis ri^bt^ i.^.^ -kL 

kanganito- 1] A1 Irlendlv b|g-[Pi1' 
^Lo/from tbe big, l^riendl> Lan^aroD' 

1.1 r3 Agreement 

Tbeftujdlku'j' fontftlns two ma jar Hrjunponetit?, tln^ basfl and IElc nominal ^rttmeat 
[JitLiis. These GomponenlE musl a^pe wLlb other parts o( Lhp Renrppine In whif'h 
chc TLllxiliai) a|?pi>aT«. Th^ri" Stv loihai^ [rt^tnttjonfi CpD Lbo Las^ wbirli miut a^ee 
with tht tense fotitaineJ In rhe tisnsf e!tJ]H'jil encUtin: to the verh etem. Tlit [norc 
commoiL biiSPb are |^voii Ln table 1.^.^^ 

Tbt fii^l sentenceH (£]. provided an example of ^inmmatacaT a^reEment of clie 
aLiELlLar^' base and tbo toBsi? elenient an tbo verb. TIlp baj4' in Llki": Jnstan^^^ W44 
ihr, lltftf roctiiiti:^ \b<: teElsn or? tlif ^crb, fttp, tr^ bc iiOD-|><t£l, V^hich Ll ie. Tbe same 
fientei]ce nitered witli ihe other imperfeni tp> p bftflf-, -Jpfl^ watild tlOt be gr:imD1r%tlcalp 

tw tberr^ w<huJc[ ht ^^ itnte ehiili, So^^as).** 



^"^Th^ p^T^r dinrf nOL Ii^iuJIl tbl^ 4yr^ or int^ phnut, only llic dmrtcr brni Lvb^t ibcre ^ir 
v^FfiF number Df itTiniFtr k(^l ■■■>iai^ [otlowf^l l^j 4 ciu^rd&clied rurijn ^fiUi Uhu ^j^npc l>vct tlic cniLrc 

^^u lod la~ nr« af]amGr[ifan i^f I lie rfLlE\^' rAJir-i|i q r^r^ 
A racRpit^L^ Jul ^t Ihc ^UKibziry Lkin -"^a Ik- fnunij m j](piFi9l!i|i 

ti mifd 4va 1tii own. 

7.; 



i? perfect] VE foGJiG) 



Th* other comporerL of Ihe auxiliary is ttie nomlllJ^l ftgn^mCTlt tUtJcfr. They 
CoELlaLil p^rbcrn and nainbeT inforn^atton whiclk rriuat Agrf^ with th^ p^^inn ^ncL 
nuniher iDformatttjn o( t|i^ arftuninMlt CA^tf^ |klLT^^. Whw[i aid JiU [tibcr-Tnirli^T Lb 
prGfiGnion a noun It ina> either be iiterprftted n^ eirguUror plural: the inlOrmiition 
ill thi^ maLcKing agr^tllH^tit c^litJc dH^U-nibiiii^ ^hicli. TbLs diEtinciicm ib thown in 
(26). In (a) the Euhj^ct^ urafi^ Ls sin^uLar heca.i£Ee the BubjecL agreement cliLac- u 
nuU^deroLlnjthtnL'p^^r^^jnfc singular, Tn(lj) Ihesubj^lL \x pluri^! h«:^nbc Llattnbjcct 
hgre^niebt clJUc^ Ju^ J^^ncli]^ thjid-peT^oQ^ plmd. 

nuD-EZG -3^ spear-PAi^T kan^area- U Tf A L 
'TtiC mift S[HZ4Lrcd the Ivo lan^^rcch ' 

nuiD-Eill G -Jll^E-LlnSo B p ear- f AST kangaroo- [» U M 
^The (i^vflr^] men Spj^itH^I thi.- Ihvo kar^roca/ 

1-1.4 Auxiliary Pa'^ilioiiing 

The &U>[illar J v^Orrl tnusl Hippear either at bhe b^i^inninp^ of a EenLence or ai3 W-ofler- 
nbgel^E pD&itionfU'fljz^^], L|bf "'flH^rord^ po3iU&n. More prttianily, tEni second ptailion 
HCUr^ ^Lt lilt end cf the fiiEl phenolDginaJ phra."^ ef Lhp Renternre £it in th^ second 
phTaBe by LtEelf. The &qxjllAry may either Ise t ivonl unto Hw^lf or ipp^?^^^ as a clitk 
on Kh*^ last i^ord i>f the pbraBe. 

There are raoce r^iie tra] iit^ OJb tht pu*ltioiiL]LE cJ LhJ5 iujuh^iy. In Warlpirl, 
words must hftv*: twcnfit rtit^reBylljlilcfi. Tlberer^jre, if the auiiliaiy ha^ onL_vone overL 
^yllitblHin fl.?.i ioj then It inu&L he ermlitic; J^eoce It must appear in Wa^lwrnafttl^i 
po^thon^ n^tber than al the ho^nningf^f af^entei^Cf- Th^rv i^ otii>e>;roptpr^n i^hkcll lu 
the auj^illar^ h^^^ ^^, ^'Ihkb l.^i ^iLlJtEEs and [iiit> not begin a wojd eviMi LI tt be^nB 
an Aiq^NilllAry with f^o or TTiore Eyllables. NoLe LhaL the agreement markers are also 
clilks and also may not beplri a word |t,^n 111 tlic OVJ^nt ol a p honolH>glc*ll v null 
haae). 

The-hentem^es in f 27) Elr^nnillnl^ gr^uihinBlic^al pl^emcntF orihi> auxiliE^Ty. In 
(at the leEt-mcet element of the ^tjiubary^ ^^w-Hfl^ m ati aftreement mftiker fl-nd 



LberbifDrE ]L musl bu ejucLLjc to tb^' pr^cEclin^ word. In (b)s the aiLxiliikr^ ^uaEisLfi 
□I ^ ffingk sytlnblej io it tt^o nuaM he <^tiflltbf atitl hi WftrkpT-nae^re [HWlLiriji. (c) 

afLwo words. Jn {dj u Aiujliir^' iu liT^t pceiLion j& eIiowil. 

^I wlLL iaLe LLe boomeriii]^ TroQi LhE cblld^' 

rhild-lkfrtllF iiEE-rfDN?A^~]' miLn-Enc 

"I'h^ ni^n SDK lJle cbLId-' 
c, ^fa^fli ii'(n-ti- rTj*T-rfn t^jrJ] pirnJn-rpJi -rySU}islH-T"liir 

baJlg[UW Ii]^-P,^J^-|e-^ bocmiomii^ LakE-NON1'AST l-EHG 

'I ^'ill CatE tbo bDcmcran^ Imcn Lh^ bij; kau^rocL^ 
h]. ft"oJftfrrt-tijKi-rJa t'Jirfi tifj-rfu^^ ^jLj^^a-mt 

ADMOPf-:&-,td bocniErarR chald-DAT talie-h"OPfPA5T 

^YdU lELEgbt t^hn.' |]|L' buUClLCTilElg ft4jj]| K\\K J^hM.* 

THe ECU tf nir^s in (^] E^r^' ilq^ gTHLnimii^lcal. In {a^ tbc anShll^ry, ni4f-r/a, ApJ^i^lr^ 
j^tUHc to tlni word In llio fr^uakcl |fclLrM<'- L]^ [h] the clitic, rJa, b^ina z. wurd- And 
lEL (c) rhE auxjlLar^ ^ppcdr? f.li P.icE^h'rl la tbe ^nrd in th^ tliJrd j^IitAm^ LtiD 1^1 tb^t 
It Is enclitic lo tlie verb m^Lrti nn dllTfrenct^- 

J'^^O tfLk^NONP^^l'-lH-^ clhild-HAT baom^riDA 

3d child-DAT LakE-NONPAST bcwuiEiang niaji-E.JLG 
c. *Xtart\A cuiri-Aj i^ni'/i ^rjn^a-mn-mn-rJfl n^jw/ij- riu. 

kaii^Aroo bLg-UAT bc^^mk^t^ng taho-Mi^SPA.'iT-U^Sd l-i!tit3 

1.1,5 Ar^nieiit IdentiBcaliou 

Tbcrt Ls ^u LLniiortuit rtlataort l>eliHT>ii thn verb ami tKe n^aee phn^ea In a simple 
i^^nE^ncE, pAm^ly, ibf^ rpLation of prcdicaiicn. ThaC is. Lhe vErb acta lakE a loj^ical 
prediEaLar^ Lokin^ Ihec^e pbr^c? :ui it:? ^rgciniEhts. TbliS reJ^ilir^ll Is itL^nlEitEt'd lit 
tv^ iv^ySn Sy nia^' Lhi^FLlly, c^£p pbrafi^^ may appe^j rb liks £Libj[?rl of a £cnU?i]ce^ be 
ft'ell BE A diiEct obJETt aiir] indiiEct obj^ii. Tbi:^ 31^ dlf:tinct from the semantic it-ne 
nfraiif pbrFUi^fi in wlkicb tbi^y arc idsntiliHl whb IbE diir^Tsnl reins wbkb 1be vctl> 
BJ?lEctB. Left clar]r> Ou^ tw<i-[cVHil aiialyMi t>y CH>nilrir'riHfi [1) cifli^f e^d. In thU 
^nLEncE, tbc cr^LL^'cLy markEd prai]oun^ n^_rrJu^ takes dn Ibci fitlb.itct futif tLdH ^ 
inrJ?, this noijh markeni iot ^hsoluMvf^CJiSfT tak« nn tb*^ object function- and turJUf 
mftrkEd Eor dati^E ra^E, Uketl on tkc Jntlir^TJ ^bjftit fiih-?Ui>ll. Fr^Jl t-iie yetM-nHc 
pcflLt 4f vIe'W iT^ii^ LIiaL ugAijuiM iR l|i^ is-kpf, ^h^t korSiih Lbe llhing taken, and tbat 
Lurdu ih tbc ftcmn-H.- frrjra wbkt tlir objetH i» tillfli. It is Lmp<>rt.aciT Tbftt Lhf pwvicr 
be ab[E Co detErniij]€ ibts mappLni; From cbec pbrases to a t^ulnch t^. slho^iL Id [29), 
to puj-t oft be mH.-a]iijiii oftbi^^ejit^nw. 

21 



(2n} taker ^ nyjjEJ^ ['I') 

tfllltti — iarft (*ljoornerHi3fi') 

1.1.0 Null Anaphora 

The laat phefionn^pon to hi- Ctavertfl I5 kno^'n m ''i]u11 anaphcuB." In Warlpirit case 
EhEir4fe arguutCuLs hceh] nui appEiar ovcrLly m a sentence. Wlien tbk happens, tli? 
relereDt of the niksflim^ argument In r^tncvttl [r^im COdtcjtL Snt^iiGsE, tm- i?xaniplof 
ML^t tEie ^EH^bri^t ]fc^ (hix-[1 talking abadt hi^ son wben Iie uttered (^D), wtiinh ee the 
B {1} wUh iTjrJtj-tv mawing. 



{3U) ^'jjQj'ii/L-rJu fi^rflja-Wo p(+nrfl-?-ft( lEwK^t. 

l-ETtG IHPErtF-iB-ad Lake-NOUPAST toomerttng 
'I am takiiag the hi>i>m4}t^ijg Erucci hi[[L/h4jr/it/ 

Tlih^ B^tirncc w?i]]d be understciod oe relerrin^ Lo the Epeaker'a s^n^ hs in "3 
am UJiaDfi the biHHneraisR Irom my Ron/ Nt>t^ tlis-t rot &ny Iffrrerl USftJ ho UEtd 
Lm^^U^ Lt ]i]U4t $li]] DjgisUir ^Llh Lhe aj^re<ment chtk 3D the aaxilhary, tn thiE caEe 
third-peTECD abigular.*^^ 

ThiBaecticiD h^a prEaented the p|ienom^ti5- ttiiii the ]aaKer L^^iii hE^mlLOr HoweveTp 
it remainn to aporJly (hf pafSW itsplf, bM?joiid tin: very brief over^Hew gLven sailjer. 
In nrder Kg nnderBLiLnd linw thi^ parser ty yl^ufitf^f in ttL-^fldfjbce oriiatunl larL^Liae^ 
prof ^^1 ngp it ■^jll Jlnt be ]tecE]«Haj^ Id oulliiie Ihe melhodclDg^ of Ihe researcIiL 



1^2 The MeLkudolugy uf N-atural CnmputatLou 

The model of p:i|^|]Lg l^Nljj^^^ in this thi^Ui falLb iilLu the thecretii:^ fr^jnei^ork 
of naiuia] cQinpiitaijfiii. In thi5 *r>P"^'^t" there ire (our tottipoinitila: tlat abrit^a^t 
roil"! pul^tmbi^l thet]ry, Ihr repreflcnLataon and al^rithixiT the ltd pic m entaLiDii ^ and 
the te^t. The nrsl lEirtC] ^^l^^ment^ uf rhis ayathetjc mi?t]Lod olug}! are dezcribcd by 
Ma^rL^^a^«:i][ 

At one extt^incs thn top le^ol, ]e the abfltract EarnpiitBtional theory of 
the devke^ in which the perfcirmarL4:.4^ nE ihe dt^vk?^ la fhanrtfriztid *s ^ 
Zt\Ap'pLiig rrchcn one kind of Enfofj]]atjoFi io another^ the abEtracL proper- 
tjp5 of ibis mapping aro dfliiterl |i rjxl-'^el j^ AJtd IL^ i^ppfopriateiLCES and 
ader|i[4c> lur the La^k at hand are denionitrated Ijh the nenLer ia the 
choke oliepresentatlon fur th? inp^Lt ajirf <i«tEHUtirtc| t^eft!60rithm to be 
U^^ to tr^hfcilc'rm one into the other. And at the other extreme are th? 
dfiUil^cf how the alp^rithm and TeprfeentaLion are realized phy&LLally^ 
t&e detailed rompfcJtt-j- Hh? h Itt-tl U l^n ^ti li>^r^'^^tk- f|Jp. 24-i\ 



^\\\*•^' L|ii> ]bLr-JT li.hnilJi'H hut ^hiii' Fk M^^ iLq I Ml ^ liiih , ■ I^h kni^ ^r^^i 'ihi'^L hnlh Hull aBLi^ilifir^ ia 
■LiL arLunlJ^ jhcrrurniMF FrhIv-^Ij Lhr- |»Hi^r hMiijJx Klll7tt^ t>r iinn-iJ^i-KL Hi^aHhi-niii 



The IflEl H^ltmpiil of ilif r*ttjral comp^ilatioiL approach- Ln^litif; tfit iJTi|>l^tn*^IL" 
l^tloili is ti«4Eaajy Lo pn^^id'' f H>rHiH: tl^^ i- fr^'il li jn" l for Llni firai Ihre^ Glcinpiil^. In 
atltliUHJli to t>rovSJ]nfl Ihp ih&lmrl coinpTiU1]0]ial Iheory, tlLt^ reprefientaiioiL ^uA 
tJjontlLnb, and the LinplE-inF'ihtA^ic^li, PJie ninflt argye thaL thoy ara [^tlLrul lu ulio 
aaaoUcor — that il^e jJgori lIluIi ]ii facti cornpUlCtl tite Jll&pplrig of the compu tatiDnal 
Ihtory, and that ihc implc-tiirntAllon Ls *i n^firP^L rcaJizatlon of the *l£orithin. 0]KC 
Ihls M iJone^ the lesL of the implcm-Jl] tjtknj C*n he Jiaid to be a proper test of aD 
Ihree conipoiieiLlH. e^ptJ^ially the tnni p^lt alion al thecry. 

How can the metbodolo^ caf iiatiJrS-l i:<hm p u titioT^ he applied Ic the prOhltMn 
aL 'kznd'f Afi for tk*^ (IrFt. parii GD will ho utpd as thJj tomput^tjona! theory- QQ 
delll]** thi; l]]ap]jJ[1g? bel^-oeis the beaLeciLC iticl L|lc -svilt-i/tcr ElnicLnres underlying 
it. flecAUBe Lhe aiin i» to LnjUd ft parser, ^'o will he coinputLu^ lh\l itlippiriftS Id 
the directioEl Etflni sotltejire Lo ^tructuie. The qgiL Stefp of the ^mktlon, then, Li to 
deti^ Iht ;il£Orlt.lhin and ropreF-ealaLLon lU^l n^tnpnte ih^ cnappiiii^ After this tliE> 
dt-ijgn must be implenn']i1<^l. ^^\4^ lin^lly, tested on a ujluril laiLgUftatn W&rlpiri- 

As HOt^'d above, it ie nHTte^s^ry lo ptodNLCe a-lfionllims and represealatLons tltit 
are ^tbful to the i:om]Ji|t.it|nTia| iheory. To thii eudf 1 wjH enj]>[<iy t|ie lype Iran^ 
p(Lf^ucy SypothesiE^ a^ de^frlbr^d L]l [BWe4]^ 

... thJ2 conditioiL Ihftt the logN'al orgajiiaalioa of the nilc^ and &tnn:t*irw 
Iticorporaled tn apammaxbe DtimTiOd ffttllC t" e^iactly Jn the oigaiuzitJim 
of the parbLn^ ELl^orithi^i. [p. A3] 

Of cotirSei dot aLL itgtJTltliinH ner^ he con&trLtted EodiirtQy frOJil tlif ^OJlipMt^- 
tiwial tliBory- This hypothesiB ih *p[H:.tlin^ hKatjpe It mlninti^fifl the ar^umentatLoli 
osded in order to ^]ww that (he algt^rlthm ia l^tMul. (AdditfnMsAl, ji^dop^il<lBnt 
f^iipport for the hypotlieyiN as givtJi ill [JIWB^]-) Hy allowing that the al^^riLhra and 
Implementatioii mirror tfk^ Efaitian:*r ^ d<^dinH^ by GH flo5t'ly_ I hopL- tu Nhoi^' that 
the EolutLoti pror^trlj' nncw^rs L]he qnestioiu put foitii in tiiC thesiSr 



1.3 The Abstract Com pit taiioiial Theory 

This Bection pnHciitS a hrlf^f dewirtptlon of the parser it tin: iuvol of ^l^tr^r^t com- 
piqtalStjnal theory- The theory is (uiJy prrrf^Htftd iia the followiuR chapter. 

The tftjft of the abstract compnt^rtii^tlftl thn^^ry of the parser is to specify tiie 
nj;ippJ[L( of thfl Input BonLeiicB to the oatpnt ^ trrtctti ro^- The parser aRBumes chat 
Bome p rocessin j( of thf kapu I spiitFTice(i_CHj speech strtim) hOJ hcCrt pwfyrjned- The 
Input to the parser can be chanic(ci12c>d nX 'A four-tlerJ^ ElrucLtire. At the top levd 
is Ui0 5E^litH*n^e to he f>af6ed_ SeDtencea ccrhi^t of ft dnilih^r ol phono1o|^cal phraseEE 
phTass consist of ft n d in EN^r c^f i^^ixh; flqcI words consist oE^numh^^ of HLOfj^ti^ltl^Vr 
A* *n examplPn 11) is j^V€n belo-^^ in the input rtpresfntation: 

(31) {((IKiJtJLU ELOl (WA RBA ULiIJ] CCPUITi ^l)} ((KUSDU HtJ)l 
UKAFLLD)) 



K 



TtLestrtKluTflof tlienutjjut Jh gi^ci] by two Unsulfttk thrarieE, GH ajicl a HieoPy 
of «bt Ib^icoil^^ One t^f thE goalE uf Gli b to icconiit lor Clngujstic hhsDOEriEiia 
[inotLly EVDtajlk) A'ith j rumber of levels ol rej^rcsen 1b [ bn r i^Mli of thrae levels i^ 
toii^<^rtie^] wilhacrrtam .LSpcrl crit^e linguistic inCcHrrriaUoD ciJiiLamftlifiHi Bftnteiice- 
TIlg id^i is tjiit h-.^th JeveL reprpqenLs ocdy i^hat it ne^is to Ln order to mroiint 
for tllAt a^pecl wiLji wliirh \l IB CMcerned^ fltliOr levels repr€fl&nt tJie iiirormitiop 
appropriate to tlielt liomains- 

ThiB apprnacJL diflcrs Bub&tanlially ftflm earfiGr formuliitloilS of ciatniaf flizm- 
mir&. T fuse ^im nmrs r^n^littd uf a si iiRle BeL of ru lo^ naiiJ I o Eei]eraL* &lruf (tlHM 
corresponding to iurfitte fitringt- \ string witt haid to be gramm^tirlvl Lflt noiild 
be generated by Ihe grammar. iJld uDE^ammatical ll jbot. TIlg CB-5tyLe appTMnlh, 
on ihe otiiflr lllind, uhes fieveral atructiire* lo seneratp frurffLM Sirin^f bjt each b 
LOnceji]€d only with :ic?c[i€ isp«t oi thc^ -^^ffitencr. Bec^uBe more thafl urn? ?tru?;tiirs 
is iHftI, A sflnteace may h^ paJti£jJy firam mitical. That I5, it ina> bJ2 j^mmitical 
with rei^p^t to some asports dT tike grAniJI^^f, and unRram m a tK il ivitli rrapect !□ 
Others. ThLB formnJatlon \\.ds an intuitive app^. ^loj^eid^r (32). Formally gpn^ak- 
ing, tills iH^nlence is u n gr^m rMlIf ;ih yet iH lb nncteivUlldihlo. [f only weni anrf J 
were inteTrlbangj.»d, the sentence woniM bo Crtcnpletely gramniatKal. Jioughly tpeak- 
IHE, GD would rejjli^Ell the paiti-il griitnttlEttii,itty by Btiting t|i:tt (^2) hft* an 
nngrammatLCiJ precedennze stru^ittiro, bnt a gramm^Ucal fynt^tic me (i-S-, that 
CMe-marklng iTOJld Ltionji are vLoJatfd], TikQ Eriminitjcal syntactic Stniclnre illowB 
semantic lElterpr€tatLQnr 

(32) Waiitlto the store with Mary. 

The venioil ol GB adopted In this Ihtrais com^ laigdy ffottl the mainstream 
work 111 ttiH^ Reld. Ilo^'evfj-H ttn?re are many iHjtri&tKfiib o( the ibeoiy extant, mostly 
dne to the yollt^ of tbe endeavor. Aft a ros-nit, it draws from ftc>jn<- oribc GD work 
specifically focused on ^Varlpih and tijuiLax langgagns, Tln^ ccLaJor difleTencefl here 
coftccm the fonnulatlf^i u( tbe aubiect Rram m^t ii^al rdaLiojki aud itae of itit ^UiC 
category AUX infltead of the morfl common INFL fLniTecHQn). Tilt foLfowin^ chapter 
on the jtfigutttlt theory will ps^ijit mit where tbt twii at'^^ounts diverge. 

In the developniejit of the parser | iLa^"* [iaJ tn miLke a furthn^r JnodHitaliOiL to 
ihe nndertyJngjLhtoTy to aicoiint [or [r«e word ordtTr Therfl Livt been proposals In 
ibellteialure.^' i^jt uure seemed tt> -irork uUt ard bLlIE maintain a fi>n-^0rd4L]i(.e with 
f-H Thi^ cllaci^ adopted in tliifc: tliesis bi Lhat 4 ^truttnirt and the other syntactic 
slruciuTM are Linyrdcredf icaiTDg thtjii eoJcly e^ hierarchical ent|tjffl_ This tOu will 
be elaboratHMl lii the following chapter. 

GB is composed <fcf three majn Le^-elfl i^r reprosentationi phonological, Ic^lcalp 
and syntartlc. Tiie phonofogica] h'^\'\ h meant to capture t|ift iw^Mtid SinKtolr^ of 
utterajirti, while the syitacii^ \^vii[ representB ihe ftyntactJt roliiliouK among Lbe 
crjiifltitueDts that t|n^ pfionological level bru hl^bllgbtcd. The logical level, while 

th^wf ar^ ni*ny [h™>rici lii^t fomt \\\\^^'t Lhe porvicTT [>ritiiul throry, ^ml thf r« ate c^rrr- 
DTNiMltaKly mahji j*ri|lh-r^ h R*>r-J rhciMA- ia [Li-^Ho] aji4l th^ rcbn-iiru mrHtifHi^ii mt^irt, 
~^|t!hin1Sl EE A rw\s.\QW vnifk fiT Lhu era ^r trAahr^ridi^idHr^h lip^LiE-.i|i^ 
5ft INaaWjaniJ Lhc litu-itfinum rpj othtr Lficunes llitp-- 

?7 






B npi^SKry part of a anore ncarty rompfi^tf^ flrammsiT and blfthly (l*^elt)[W] 111 <^S 

tuQii^ uniic plinhtilrtiytal .irinl hyDtKi<tLC To els of rT|>t'JJ=sHlltiliuEi, A gnkphjc diijjicLjDa 
of thi^ pn>ce£5]nj^ ia ^^gil LeI figure I-B+ 



Itput aFTilppra 






SS 



Fignra LS. Tlir par*lii^ procWflr 



In GB IhL^ Lcvd of plkQFi&lopcaf structure is caHn?] TliOrttalog*cal Form" C*^')- 
PF ivpreccDts many ^l>Wt^ r>f a -Rf^ntcnctT ancb as pilch, fitm^ft, afitl nn^ter. Only 
a p*rt ot i'^\ nanielYi 7- rtcnidHjm:*- ^nd srlj^^cency TdaEians, are ased lo form tho 
bajiE of parser's pTri^nHfnce Ht|uciure_ PS rcprW^lt liils J nfor m^Ll'in Iw all levdj 
uftllC inputs morphemEat wotdfl. and pl^rilWftr By rftprttflenlmg tt^ precedeDce and 
adjacency [Ur the i^Oipli^rrWS ftf a word, VS aLo J^hcodra: HOrtLC ninrplinl ngy; yet, Jt 
cumot bf railed i morpholoj^caJ blrdCltlt^, as it ^ko dealB witi] lEnits tarter llhan ^ 

1'he phenotncDd arcfinntcn] for bj 1*S &re givfn m tabLe L.3.^* Tht COtnpOSltiM 
of nonnfi^ \i*tbi, and AUXili=^rtFi u cov€red by ordering atTiong mor[Hheniej:- Word 
oHn^r B used to accoQiil for contLllncms <^^ plirasea- Fmariy, aiujliiLTj" posLlioiLing 
Ls CQUcem^d witli Otdoring ir^f a wtud lor clitic) *tii]ODg thu^ jihrafrffi of a teritencej 
fipHSilkalljp tbe liraL oraecond pb/ftM. 



^ iSOmiltaL verbal, and au^ilLiU'}! composition 
^ continuooE taiC pliJ^W 
c au>:iltary positionLiLE^ 

Taile 1.3r Tho pbeflOlrtena rapreMoted in PS. 



Tho ft>ntMtlc |pvel IE further broken dtiWIl IJhtO two *.jb|evels_ S-strui-tutJS atid 
D^stiurturHi, that are ineanl lOCAJiUir^ iheayni-witic regukritios Ihit appear *t br>t|| 
ti fiilpcrllckl ftnd dco|i level oraualjalt. S-BlrULttlffi OJiT^spoodfr 1 nore c]DF-nly U\ Hit 






K 



iLjrr^C^ UtterfLhC^, ilE^kJOl^ GK^lLckt lll« ^ynt^i^tr r^ktlon^ O^ Itfl camponents. such 
as ^(jibs^ i]oiiii£, and tike Laxj^r cDnfci^ilLJ4>liCa, V£rb and n^aH p^Lfa«^. D-£tructi^e 
repr^s^nU tbc nai^H^nlf'-^l fnrrri nf Lh^ BPntemre^ ^liich nia.y have several Burface maA- 
ifeEtaLiQiia. A qLiicL example weU brin^ our the d]sliacLiGn. Considet th^ i^Uv^' Uld 
pMfllvt forma of a s^^tittnceT *uch »* In [33). Thwe e^terw? appear quite different 
when Epoken (or written), but they ^eem to ^ny the same thin^ Th^ diEFer^iu^^ In 
appearance wnnLd bf captured m difF^rfcT^g S-^tTucturefl fof each of Lhe ^enteuces^ 
UoViwvr^ Ihn EimilanLj wuuld be captuied with identical D-BtnicLuTii« for e^i^h Si^n- 

{S3) a. [ tooL tbe boomerang from tbn? thilJ. 

b. Tbe boomef^ng was tA^«n frc^m tb« child by ^le^ 

S-atniCturc ^T^id D-^trnctiireaTe related by tbeam^e relation of ^^i^rji^nl, Ths^t 
Ei^ elements of D-^trnEture nia> move froni tb^i orlgiiial po&ltLt^tU: tOOthf^ pc^Jtir^QB 
In the Strucluro. TIlIs will bt elFthorittwL in th^ fnllmvinj chapter. Wbat ihotlhl 
be noted here ia that for the parser Lhere ib no difference belwc^irn S^utniCturc and 
D-£tructtirc btcituw nnj HK^venient 1* nwf^iry In ajinpl^ W&ilpiri sentences. For 
thi» reftflon, the parser nf*d only t^mpule ope ayntactic Btructure (ilS)^ acid nut 
l^v-1. In mote traditiiioal GB thcoijf S- -and D-strutturc represent precedence ac 
well as tiieramhlca] inlorn^ation- Ttius, movFirient is needed to account for either 
movement in tbesnrface Etring{precedenceX or movement in tht AyntACtln: structure 
{hierwchyjT or bothr Because tlif &yntactjs^ struciuie^ adopted for the paiier are 
not ordered by precedence, no movement i;3 necessary for peimntaticm in the surface 
jtrlng- Thflfljtnpic riiigtofpheiimnenici>vefedb> the parser demand no hierarchical 
movement, BO S- 4n4 D-ftTOCturefl colU-pne Tl^ia cLoeiin^t constitute an aj^m^t 
ag^nat two l*velj o( hjntactic reprasentalioni as they 5wm X<y h^ neewsary cross- 
lin^lsticbily. Howevern tlie paj*ef nen^l OFily repKsent one level 

GB consists of a number of other subtheories^ eacli oE which define the rf^a- 
tlCHlU thlt id^ obtain at each of tbe levels, and the doUStr^lnta whicb grammatical 
Etructnrefi muF^t JiatJ^fy. TT theqryi fo7 examplei rin^Flnes tbe atTuctural pUslbilltie^ Id 
ayntajL^ it ^ves tbe ba^ic possible htructureft for verb pbras^Si nouil phra^r^r ^^^ ^^ 
on. ThtthH_^r> of povernmenl defines a widely tas^^d itrUHlturil relation, govemmenti 
that &eenu to pervade the analyai& at hotli syntactic levels. Caiie tb4?oiy Ufav^rs tll4 
osa^e of case, as for e\ankple, in tbe cas-e-marked ooun, Aurcfu-^u. The lA^t fliib- 
tlieory useri by the paittrs ft-thcory^ Ls *hont lhe eemitntie snbcategoriEations of 
predicatomr The verb bteni, pimrid^ for inatance, suhrategwiMfi for three semantic 
ar^menis^ one for the taker, one for the object taken, and Eot the source of tbe 

Synt4Lctic structure (^S) accounts for a number of pbcnomentr M given in ta- 
ble 1,4, Frte OrJtrlng of phr^seji ih j^prpsLTitcJ in SS, mostly because there are rto 
ordering i^onstraintt amonj; phraser in P£: S5 is nnabl^ to Impose any su4i1l con- 
Straintp fu pTccedence ia not represented it thai levd.^^ A& claimed In the theory^ 



pkruv^, ihp ^-41 tLCimboii T tue pbrun wiLb ileiiiI&j cuE-ninrliH^ «^uj|e] U# ^jDkfi«d La ike mawav 



A>UliTtllk:i| functlcuih arc Knirly til'-rarchiraL ri^LiUons, sc* tj^rij- LLto ue repTCBCEiLed 



Free plira^t' f Hj-der 

[) grajn]ualical IimttiorEt 

Table I.E. Tlkr plienoineiia rcprcBcnt^^ Ln SS- 

Tlif pwaer Ls aIso L^afitd i>tl a limitiKl BemanlJc tll^orv- Tht ^miintic-:^ E^Klends 

K fftf M ii3l€rpn?tatlftn H^f tliF^ Byi3L.w.tit f^lnHTt^Hii is potiblblc.^ Tlit pflJBcr ccrrera 
Tour kinds of sejuaiiOi" iniprprelaliiiDp UftlrMl in tjible 1.5. Arftumeri ifJoDtLftcaLLDa 
IE the prK^SS of rplatiiii^ ChL^c plif^vses Lfl tlacir argUftwri p€jfi]tioisa dF prcdiritw^ 
wbkh, Tor the g?piBac€fi iu Ihr clom^m of tbc pintr^ :ire ^erbs. Nail ftjiaplmr* 
Dcttir when tbere ifl n-iy f?vi?ft AignmEDt in IfiC seihteDCCi yel Ihi^e 1& *n u nd efb tood 
arptmtm, U?.ually gleaned FrOCii dincoqnw conlexl. 0i>^niee ihe parK-ir prwc*9« 
ottf MFiteiice al a tiifi^ tlio ^rfturac^cit is only li^El flngsed as reftrriJkg io =:0FieLhtn6 
outsldt tlie ?*^nt.erce_ Null nTiixiliary tomponenth arc lUt^rpietwl with iheir ddituU 
valut^H ail deflcribed J^cvh;. TIip last phenoineiLtiQp AUnillaTy agreemH^nt. ]% rith^r 
atTii^tiori^iTHi Hfce aapwit Lnform ItlJt^ Jfr n^-Juntri Eied Willi thfl teiM** and mocd on 
the verb's tense e]enn:tlt, subject tu the Lenhn: m^tHiTtiona on the h^EV^ tlifl [LOJIiJtiaJ 
a^^mcnt Lufcircnation frocil tbe*m^i(i*ry is combLncd WLt|i Lhe arguraenlB (i.c.^cMt 
phrB£^J in the sonleliCLin 



^ ^^rgLicnpnL IdBnliliralkiD 

« null acin-ph^iiV 

^ null au<ilLh.ry cciCtipon^antF 

c Lcrit JLilcE *rp5i3niF-nt n|;recnn?iit 

Table L5r Th^ pfionnmenJLinlErprElcd by MTnantiwH 

Lexlc^ theory pJovid^E ail iHlHlOif [\ t of Lh e 3 ofarcniUon A-'^SfiT t3-tfld wiLheachle^caJ 
itfiin.-^^ EiCiL jttm maps ta a lexicE^l ctitry th^L ronLains itt ratJi^ory (i-e-, pari of 
apeech)^ infoimaUon l^r h& Tole in both pretcdeJiCt aunl ayntictic stmctiirHHij ntfld 
sHicn^nHflriroTmFition. That ck iJlustJfl.ted in fi jure 1-9. Tfan? Lufonrntlon In theitem'i 
entry i* ivhat. cJ etf rmines its interactjon in tln^ atrtf'tiqrea at eich of the ItvetH, In FS^ 
for pT^iT ppU the item's i^tklzfttlnn miormation deteitnlnw what entities it cbji be 
tntlltic tOr The precedence part of Lht eiatry loT raae-inarken . for initftnCf , Inrtl4'ateft 



l^rd Gf i>DlLK An witli |>-9tTDi tnn^, I.1-" u relalrd Ld h ^irLictUK tj ininviTFtnt 1^c Hi¥i}ile p4ri 

<Lf Wf.rlpirj ib^L m ciiinjtnJ by th^ [i4rnrr iIeki diiL cill Car uij niD^vlnr^llt I ft t]i<: mLppiTii; La LF, bq 
^ Tlltr *'=rve br scinuhUc inii^rtittiUllDn u wdL 

yDi[|Dt ]?^iJ ii\ilty, 

30 



ihftl LliEjy may he Enditic to naiins: JLC^UILfl, On Ihc otliEr hand, may eldL be eni^lktlc 



lexlciJ Uf'ns 



leXJcoD 




cat^carj 



preffldeni^e 



gynt^K 



g&m«itn^* 



Thfl way LCI which lexical Iteiil* Hbo^ up in synLactJc &tructurp i^ more i:Limi>li' 
cnt«l tliiJa prJTCcdence stnqctqr«_ In PS, catli Itxital ileni ch ei]Lered as a singlE 
node in IhefltnictiirUp wllOtJ: Jt I& Enbj^t t*> Ihc lulea of comhinatlcin of p r^-^H*tite 
HtinrtLirc (^.5.. cliUclzacioD)- In SS, howev^;?, khjIi ]^^ki^ ilen] projccCs a£ a texicftl 
atnicLure { L-atru c Luh^] thai ttprH>*c[LtB che parL of the Maitentlal aynl^^ th&t ti>rre- 
Bpoiads to th&L it^^ttl. Tht fonn of Lhe L-slructuTe flepn^nds Oft tin? ]^tiilL]Lg of Ihe 
Irjsifil JN^m. Tlie theory ol ItUn^^l seanan^LCJi adopted for ihe parser diutde^ Itxi- 
ca] items lljLo tiw^ SOccUttitic claasesn predicaLors {c-j.. v^rbfl) l^nii argUnH^iltri. Tbn 
L-struvKUru [or an arfiument Ie almpJy i tl<>dt. The L-slrutture of a predLcaLor, on 
the Other hand, is a atrmrtnre that baE pcGicJouE Ln jt for argLinienlft- For e:um- 
ple^ ^^mstder (1} once aRain_ In thjs Hftntente Lh^re ire Keveral leKicaJ ileniE. The 
nn^nb^ Eince they ar^ jtot prtdlc^^Li VOn liave L-fitrucluieii th^t are ?impl>' noclw.^ 
Tlie verb, in coDLraal, is pjedl<^al|tfi?, atld rtiup maps into an L-atiiicture containing 
pofiitLons for ica argiimwit*- These L-fltrucluieE land those of l]l? olh^r part? of ib^ 
sentence, Ofcoiarse) are combined hy rule* otsyntaic to form D-slrucLnre- This will 
lie elaboraleJ m ihii: Eoilowjtip chapter. 

The Bcmanttc aspect of a lejtical entry is, perbftpi. U^ h^art Ono of the major 
liypnthes^^ of [^>jca> semantin-R 1ft ihftt thtr mt^wiing of 4tn item in l.LTge part Itzcb 
ltd S>iilacl]c manJf-i'Ci^tlonr While the work is stJU ongoinj;, a number of cojilirm- 
iiLg rnrsiiJLs have appeajetl TlLt Jjt^rsci adopts o]be of those thooricEj conceraing the 
syntartk nianlfHSUtion of arf^ meii ts foj [^refl K^tnM*. Tii^it JS, the ar^himent |K*I- 
tlons in the L-atruf tUT^ of a prcdicalor r^ed not he hpnciFed in the sjntantic part 
of the entry; thej titc, instead, determinable from Lh* *iemaitllc iii [orrftFLtlon whmh 
encodw a list of the argucneuLs ajid IJielf ty|30*. Tbotc Is a act of ruluj c\a;utcd by 



Gil pmTidc» A fuller ^T^OLinl ■:■[ DDaan pLcuci, m^iPJ^ tft^m ^ pFdjrrioff ri-i^rMf^ij r' dijE !□ l\\\iif 

fiEfrfK^Uv^ jimjrc In LtiLJ th^^n, hi^^^VVcT, HniLDS nr? (bLen tu br QfiBnd.vELblt CDLLltv A mnrc 



.M 



tiLU ]5ar5er Ibil wJll m=tp Hemajilk areumfcHli& Lq ihegrttcimacif aI J"ui]caol3s hk wtikll 

(S'l] I, Tllfi afiL^iil of Llie acliiati -ippcaitt a* llbf enhpvL. 

2. Tlie palieill of t.|ie aclioii a-ppt^flia M ihe objett. 

3. The witrce of the ACtiOil ^ppo^rs ae Ifi-^ ilitlirfH-L object. 

ConaLd^r (1) one inore time, Tlifl l^-iiii pn^dkctKu la tlie verb hL^rtL^ ^tj"^a In 
]ls leldcal Piitry ]e tho si^m^Cltin mformatLcm thfLt It takct IbtM ftjgnnientes a- tftkw, 
a taken, Mrf a sonrce IroiTi wlhicli the taken H taten. The Uker sa the afieat of 
Ihe Mt^oD^ and ivIiaL u taken ]& Lhe p:*tlcTiL ApplyuaJ ^^^ ^^^ J^ l^*^)' ^^^ parMr 
delftrtcllnftg that the t*bet wtLl appcnir a:^ li-iibjei^t, and bo ->n. lhi& wa5 shown m the 
mapping of ftr^mmaLLcal funtLLorLE to wdhIs thftt li» the arKumi^dt positioiLB La 11). 

1-4 Coniing Attractinns 

The remaining ItveK of dracri[ftbtl IIChI no in trod ucdc^tl , and k ia bft*L rH> ? imply 
pre&eni tbtil] et raJghla^ay. Thf ffJlowin^ chapter providet a cMlpletf^ deEtription 
of the Imginstk LlLH_-H>rj underlyLnR MtC f^n^J The di-^igu ol th? parser Is ^v^n Jn 
tbe chapter, HepieMQtaLlon itld AL^rittm. Bc^fiuae the implementation folL^rwa 
LLg design qoile cLc»e[yT I\o separate tliapter Is needed to <Ilb<^iie£ lL. InsLeail^ a few^ 
Implementiti™ notes are RLvt^n Ln the appendix Tlh^ appendix alsu ffprtain^ a bnt- 
lei> of EtammatLcal and iinsrn-mniatkal inplat^ jlven lo the plRer Lo d^i mons ttatd 
Llfl toi-eiaRfl ol the arivertLjed pliFznc^menfu The Lh^ifi frtn^^ludea with an rv^lnAtlom 
ofthEabortwmjnBsof thi^pitseii and acomp^nson of ibn" paner with almlarU^ork 
in the field. 



33 



Chapter 2 

The Linguistic Theory 



CaiKrnniei]t-B]t]d]ni tlioory (GBl and lei^icaL Iheury t^^imprise the rcplftMSnlationa! 
roimditlouofthe Wartpiri thWMF- These Ihtohes are bj no mi^Aili romplete or weD- 
LindeTElcKKl, Tiowever- M wLlli Otlifll finieiilidic LlLizH>ri^T they are in * cOFi^tant sLatn? 
ci[ rliLX, i^lian^nj^ rat>iJLy ** new msighLi in- mAde. In llil* nztuipter [ will fiUtO the 
partacuLix form M liiU ods oi ll*r6t theoriea thai IliC p*rapr aMiiintft. 

Both CB and Iwit^ ihcory cotnfl unrfer Lhe rubrk ol gciierativi! lidBUlstic the- 
oTj. Tbty can hp viev^ a* ar tnirintLc siit^cLJlf^rion d( tb-> grammatical sGJiltJit^ea 
of * loJiyuage, cnucb Ijko a li^kil pnrJir^to whkb ima^tititly denotes Iht m^mheia 
of tbe tet i^\ tlomentfi for wbitlk Lt would yiM LfUf- However, not any Btiltmcnt 
of iIlc pftn^TTiflr will dij. The power of CB and the IsJiJ-iRl theory ]& fheir modo- 
lariLy and n]gulflJ-|Ly thai eivt thflm an exfiliiiattary punch. Tbc theories ccmELst of 
ft small rumbcr uE con^pojipnts ibit MJnbjne lo make isowerfnl prei]ictlon& abottt 

grain ma.tjc^]t>'. 

In an attempt t^ -iJ^lsUin the iin]^"i:rtal Aj^pects of ^tammar (i.^., the leatures 
common to all Uneuagefi), the niodiilet ol Lbn.-^ Uhforiea arr fititnM in & ^izncral 
jnajin^r. Even ihuugh lar-^^^^ ^1" ^'^e wnrld seem LO K^lslhit aimiLir pbern^ena, 
thej do ED in diffi^riug wave. 5o a -SinelE theoretltAl statement cin not BnfEce to 
account lor the ^ur>in^ tlata On Lhe othfr llflnd, a LhoOty that liBLa oii^ll of the 
ca-Res serves no explanatory fnnLtifjn. (JU resolvefi lllis rllacrepancy -ttlth lhe notson 
of pat^t^nfiJcHiif I'on Tlii^ tbeoTlea within CB are Inrmulitcd iii goneral wiy, yet tlkey 
nre gnhfctL to limitL^I [larameterizaUoLL loj partkiikj lingliagw. An CX^mptp can 
he found with the phenomenon of ape^-menL- In Warlpln, both the auhiait and the 
objecl have agreement mitrkcrfl In Lhp aiiKiliarj Jr ]=:nfilisb. howc-i^r. there i^ OIkly 
Rubjecl aereCJiStQt (fojnd on the ti^ase marlzLnR oE the verb}- In fi^■nr^^al, one migJit 
BtatH? tbat th^ argumpnlb of a verb must a^ee wuh the uEnyiTH^nt marktrs. The 
^^a^aJnete^ for Warlplll would ilat^ Lhat toLh tlie SUbjeJ^L and objj^t ije Involvwip 
wtiile for Rnglish tlie i^urartietpi would be fiet to nubjecl only 

Bui tb^*e theoTLOs do mort than mersly deti^rmane membership of a s^ntrTirp 
In a [aifiuBSc, Thev also imparl liriRntHtLcally relevant Eliiictures Id the teutenctaj 
thalmaJcc thcfiyniai-llr IjifoTm a-tion cohtJlined in tli[?nn-N|>h<^lL- Onesurh Stnjctnra 
is\ Llir n^lfitiPMinhlp IhMwccii a proliisLlcir' ai|d H^ ar[^iiFiii-iil-h UmL h, th<^ IhenriFi^ 



^flji 'prtiiJCfclDC" i mran m «b^jLj iImL L*ki^ *4J^iMaEii, \ikz ^ k»fiic*1 toilttJiiB. Ofl« prttt»i-tk 



idenLLfy ibc predic^ator ^r\i\ Lie aiRuments in Xhr sCilti^jlCi^T apd Lhcn dGlermine the 
n^l atl Oil fillip B bel^'i?^!! t^LL'lr[- In aiinple s^nienca^^ LbJs iin^ii^ Af:f o r Laa nl nR which 
Word torrftsponda L* llie vrrh i^nrf wiin^h pliraacs (c-j.^ case pbriMS iu WtiLLpLrJj 
cDfrespcind lo il« syiiti^tk iLrguTncTitB. 

'J'he fii^l [DiEL|?oncill riithe paisGr^B IkgnltUc liisJft, (Jovernment-DLndiiifi thEory, 
CJ>nt*l HE several le^ clt tif n_^EHrt;*^nTAtloiiT nonce rnifip; the plioiioLH>etL' air ay ntijztic, ai]d 
Ic^cal MprctJ t^S a ^fut^DCfc This i? shown Iti llfturfl?-L- l-'ach of Ihci^ LcVH^lt cAd bo 
thciiisKt of as a dilTtttait view on the senLeDce to ^i\W\\ t-hfy forrespond- Looliani; 
tliroij^b "^^yntirtic Bun^asses" cAh h of thftsf Icvala Ultere oul th* LJiroilJjatlon In 
the EeDlercEi that rfofs pot app[y lo iLself, lotUtlft H>Slly ihe porlinenL infomiAtJHJti 
tlirOLIgll. 



aentenco 




phonal i>gi Li^ LnlcumaiiniL 



ayntactac Informalaon 



logical LfiTormaUDlL 



Fignrt Z.L= The principal levels orGB. 

Flfiian? 2.2 bIihJVj^ thfl C^B model of RT^mmar Maunietl Ijy iht pajEar_ The level 
of lo^cal repretenlalaon is JlOt sl^^^wn because the parser docs not fOTli|iUt^ lofi^rj 
BtruClU^M. Whllp the level of ^'Logicil Form^ (l-Fl Ifr aJ^ irriporbaiit componeot of 
lenterce Uiea]^!]^, il b^ nOl yH_-t hMt dw^lt with an the paner. 



seaLBJirc 



fl-' 














^-fitroctme 












D-ntrLiirHirv 











E^1gurr2.!l: Thi? GIf iiioclcil of graiannarr 



AJRLimECil^ 



34 



The levet of ^^Phcmolopcaf Fr^rm'^ {^^) Tcpreaenta th^ pbonofec^fril a£pCCt£ of 
R fleniomce, sicch u^ |iJt£li^ EtrcsE^ and meter. Howc^'cr, cnl^ it plrl uf PF U U^ed 
licrG. As 6Hti?d in LIlei LiLlrotluCtlOfLi the ptJ^irMSUmflS In^t^A^ ft level of pr««dei]ce 
fftrLiclnTf^4PS} thiAt r^prei^nts both precedence aud adjacency tel^titMlA. Ai found ljL 
the traditional level ol PF. Note (hat PS alto IdtOrpCHtitn^ SOtnc trLL>rplli?lo6yT n* It 
represents the pieced pnce and adjacency cf ibe morphemes within a- wchid. 

The dcmain cf PS co^*i^:^S i>[ rtiorjjli^^jne*^ ii^'Oftla, anrf phrjfies, as al^ of theae 
elements nia> be invoUed in precedence rela LjcrnH . In Wai^piti^ ^ It tUHu [>Ut, th^re 
art no prH:tcdtnct cOn*:i™jiti. hdwj^tl |>LiraMW, SO Wllr moriihemefi and words wLLI 
be represented in F?. En other lan^ac^, t-u^b as English , pri?dM]4_-ndO among plht^GA 
Is ImporUJit, go thuy sflll Ijh^ mAntfeit m F'S 

The Eynlactic component ik tbe hn^ar^ uE GB at; It Sl^iuls todayr A& GB b a 
triniloriditiojiaL tlivoryT the s^ynt^tLC level Jft i;onipos«l ^( iwo ptrta, tbe ba^e and 
a set of tTanElbr[nation&. Tbe bu? is tlic ^et drstrmilutt^ tlhaL ^orrn^pOTld ti> Lhiz 
CAm^nieal forrtl of l^ei^tern'e!^^ LhLs ts repres^nLed in D-Einicture. The &^t of traiLl:14f- 
matLOCLE can be appJhed to D-ftrnctuT? toy^eld hu rl^ciit r Uc^t u tes (S gLmctDreB] tlhat 
c arreii p and to suj^fe ?eTiten4'efl , ^he eenlEncea that wi? actually utter. Tbl& Tomiftt 
oFFerE a peT^pLcuons representation for cipturii]^ both thi' EJlnil^Nty i^nd di^pArlCy 
a[ dllTereTit Hjintftt^tir ro^itTuctQ One example of ttus phenomenon was mentioded 
in the introduction, Damety^ paiis of sentenceg Ln acLlv^j and p^scvo voi^o. GB ui' 
^iy^^ rhftse sentWlCW i* having similar U'etmrtures smce tbej seem to ha^^e tlie 
same struct u re at a deeper level. The alternation in surface form ifi J~ofl4^tod in tti& 
dllEering S-struoturw th^t GB wsiftn^ to the dlfTei^nt vtwres. 

CD claims that there ie a connection betw^n PS and the kvel L>f syrta>r; that 
]£, that there are condtllonB that impose mutual constrajut^ betw4^n tho l^voL^^ 
Thu": LA f^Aslly nJemc'tl^tt^te^l Con^pd^r tj^e ^nterniee in (E) The reEatiou betivecu 
precedence and E>ntax is syEtemattc: tha subject \s the iiivt ]hoc]li phrA4<p and the 
object ]s the secor^d, 

(1) a. Johnliies Mary. 

hr Mary lik^^ Jolm. 

As far M tho [larsoF is foncflrned th<- cotin^tii>n t^etween PS and s>ntax e^j^tt 
exactly where the surface ord er of conF^titu en ts bas an effect ou tbt ayht^CtJc lUialy^l&r 
For the Eubset of Warlpiri covered, this concerrs only tho relation i^Clwei^lL :i raae- 
m&rk«r^nd tN^ nouns ov^r whu^li it ha* scope (i ** ^ thcwq^ nouns to its left within thfl 
phonolo^ra] phrase). En Warlpiri, case-roarkerE must be sulFixed Ui llOltn^ Wbe]) 
^ noiqri and a. c^-^e- mfi.rket A.re In ^kjnh ^ vojhtiguratkon^ the uoud is Identilied ^ a 
fl>ntact]c argument of tbe case-marker Convei^olj, the nocn l]Lil irgu moJl t <if a rasc- 
marlr^r wj|] appear In PS ^j* * noun with the r^se-niarher BjfTiMJ ont<] it^ and in no 
other way. En thts instance the precedence and s>nLa^t]c structun^ ^^ in a highly 
cchja&tTained ^ one-to-one TclatiortFihlp 

At \f\.\^ point I [[Liu-t repeat a caveat EtleiltiOEli^d |J] tllf- Introdnction. The par- 
ticular formulation of Qti presented heiM^ Ie actually an amalgam nf thrrc qonrre!^^ 
Mostly, tt rcimes invrn m^nfrtrH^m ^JB, Imt theit are sonic partB that lind their rcNilJ 
In the literature c^f Warlpiri hn^li^tics. 'J'he thin! tourc i^ tin' ^-et of rLuidllicalianh 

36 



to GB theory roniid in thtfl ihrals. Tho iniLii ccniiibulloil horo i& tlie rt^nval of 
prcced^siLce from GynturUr Htinrturcic- An-Mlwr n^flnlidbnliuu is tin? [orniiilaLLon of 
prw^derre ttnirTure^ which burhji^-ft Ironn two ccior^ tfaditicmal n-pr^-senl^tif^ci-R of 
QH, PF ind m^rpholoEJ- 1 ^l" P^^L out Hi? diller€nci?a FFr^tn matiMtream GB M 
tlie> luifiE. 

ThftBHKOnH pari oftlie llnguMtic Ibnhclatrcjri, lexical theurj", a^ld^ two more cOicL- 
poneftt^ to the mHjdd olgrajiimar assumed b^- tbeparti^J-r »Ath<™'n in rifiart 2.IS- Tlie 
lexjcoQ ifi Th^ Tnapping Ij^wwn morph gicHH Wfc*l PS and sytltitx. AaBociatGd with 
Fftrh marpheme iil |jm"«lei]ce and avnt-ax i □ formAUon. I^Iip pretedn^ntt^ IrtfoTmation 
deGeiiniiHsliowtbeTnrir]5h*;mei6n]Ei.iiirHSt In l*fi- tinularly, theflyoLactk Infcrcnitlon 
delernuilEfl iu syallartir manil*!£Utli>U, Cafie-marieR^ lor exampLn?. must h* enclitic 
[i-t.^afllxiii]) lononriB aL tll*^ levelofPS, "but At Uie kvd i>t$ynta:c there jb no ftUth 
^equj^emeIlt^ ^ eli Li dziLjcm is not relevant t^t LliaL Level. Jns1*^rL the Late-mark^rr 
3e LIlO tiead IceiDtralcli^meiit) of Ita piLf^^-, laltinj; nOUtlS JU Its ajftuiuenUsr 



flCntFpce 




PS 


1 
i 


















B-slmctiire 








1 








D-atnictup? 








1 








L-^irnctnie 















Fignre5.;t: TIlc paitftr^H model of ^r^mmar 



Syntactic manilwUCitHl ofleiical ItciUS 1ft carried Out Vta le^cacal atmctiirw [L- 
Htructurefl} L-&lriJCliqrefl are Lin: ayntafiic etrnttiirw th»t corxtEpOaid 10 a ainele 
lexical item, L-atnicLirC^ ftre combing to pHwIuce the D-Htfurtjre for the tnitjtt 
iwntpnrE- ForeJLiUl]|>kLoii£tderi*efltepc*mtlia tfanaltlveverb. Rough ly -=:peikiiiEH 
there aru tliJ-^ lexical itemft, n*rneiy, ihe verb, and Its two nOin^ arg^nientt. Each erf 
t[i«^ itemE is miillff&t in syntax: Man r-HtmciurGr Tliflir Ij-fitnicturi^flwrQinbLncd 
with ByntatUi: rdatlon^ to Eoritl the D-fltrnaun^ for the cntitx; sentence. 

Heforefini»hl]16 thiaoveri-it^W r>rtlie parser's ^mmmar, one mOrf pwpt needs to 
be m*dfi, coQcerniilg Ute form of a stntf nice. Thp li.^tf units ofinalysJs are taken 
to be motphpn^es, ratlw^r thatl^ -'^^y, more L-lcmental unit* tUth 3ft pTion^raoa. It Ifl 
further aaaumtd th^t BeEitencw COMtSt of Be^erai JtveE^^ words, whith flJ* sequence! 
t>r mChrpSemeSn phrMnis^ which arc flcquemeB of phi^ffl; ard, Benlflncts^ wh-\€k aia 
Jiequence4] oE phr 



The remaicdEf oT the l:h^^pt^!^ iV[[l tJe^frlbp lhe model of ^immarj dtpktHjd LiL 
figure 2-3r In greater detajl. Fiisly ihele'-'cL olPS will hp pieflented- A/ier tlus i^ llie 
Mction Oil vynUVp tbiAt will i^ver each oi lhe F-ublevd r , L-&i run: Lij fft, D-fltrucLurei 
and S ^tri^cture- The foLrmv]iiK sL^liOEl i]I^CilB&i"S th*^ iheorv of the kxicon. The 
chapter condudc^ wLcJh a d^crlpklon of the theory orsctdailCk iat&rpret&lli?Jl hucd 
by the pwger 

2*1 Precedent*; Structure 

This secl]i?n begias with adMfripHiM i>lprfMiedenee strut ture. Tht thj^rj" wilt thep 
be ippUed Lo Ihw phouomeft* to be ateoinited for with thji^ ilruttiir*; ie^ lable ^-1. 
The llkeory wwlci EtrajghcroRvardLy for b&th tiominal and verbal coiiipi>sitn>D. [ind 
coTitLnnoua case phriSHS. Howf^r. thi^ a-uxillaiy is a Bt.rang(> H_-t]CLtyT ftfld e pot 
titpturtd &i> ne^Llv- T]"e aeclioa ends wLlh Lhe cXteftJJQIW to the theoiy necesMry 
to handle aiL]dlEary compoE] Irion AOd plac^ltl^ntn 

t nominal Wld i-erbal corapoaitiQa 

D coDLinuQut caen^ t'l^^'^^^^ 
o auKlliiir}' i^cprriEMiftLLiDp 
o aci](il[ary po£] tioning 

Table 2.1: TiL« [Lb^mn^mena a4:4;ounle(l for i^ TS- 

The elen]ei]lal imiLs ol PSn nhOr|ahtmf=«, are rnriibiiied by ml^ LiLtO larger etrnc- 
tiires- Pf> i^ renuKlve in that the resulting tttil- iiLf^^ niay Ir turn be combiiLt^ by 
HktflHi SftJIse rulw (art fXiAii^^iLe will be jyver bj^low). T^l^ rulra are constraliiffl] to 
Dpeiale only oa aiijaceat elemenL^ of the ^tnictiLre^ ordered by prJ^-^odHintft 

Ea<^h of thf ekn^ental nnits It labeled vlth Its tateftnryr aa catefiorlEtl inlormft- 
tlon [9 needed in TS. Cofttidor ttn? SAH^ple^seriterice from the irit rtwl Ut U cm , repeated 
lieje at (2). The PS lor vhii lenUTitc b^>(ore having applied an^ ruletl Itl sIlUVll m 
Ggnn-a.!^ 

(2J jV^itjN JlE- j'itj tji-rtifl-rJn punJa-mi turjJu-tif inrfpr 

I-EIG ]UPE[Lr-l6-aD lale-TJrPivPAST Hlhlld L>A'[" boomerang 
'1 ftm taking the boameianj; from rite chdd.' 

The l>^k rtllt for PS i& coTnhlDation^ given in (31."^ ThlJi FpLe alLowt any n^nle 
tp rombine with any other node. HoweV^tr^ thflre are empirical rectrictlon^ or com- 
blnatlon- Th^e reqtrictionB are captured by four in ttfi^rtl n J parameters. The lirat 



*T^ iRhdn lor BaHDi^ cuE-[iiukcr4. vwb hIcehui Ihd l^W^- fUn\^a\M tic '>J', 'C , "V. iftJ 't'r 
i^cciLiclj Tlie L4illHrr rtiJinuLh uf fti^T ortioml cuTispiintn Ci, the hibfi, tlw 5^1] |«t lad ab|tfl 
ajtriTinrnL rlihiui. 4iir| llirnLnlivn- rr^Lmmfl ■ litk. UEwlrtl . 'H\ "S", "O', nMS 'D', mipnrLfc-Htl^ 

"[■|m- h.li.Lh Fur L|ii- fihli-f>r^ianliiruiLHhii .in-l lln |m r L.mh of i-a ru I hmi jH? tflkn^ll rinm fli.i|iL^r»* 

:; jhJ t L.r[?f[LiJitt|. Na>fa awn ik^ nollr I n h-ji*hx,^i--ifJ ^i^iiiiTnii ia rriiretunlL U\f r..j*KLraiuL^ Lif 

ib]piiLJMI4tH>lt. TLr ffn-ta n|k|iiD prr#«ilL^ fac^c u Ini^^ n tftirnul^Him ^1 kia1hi>«r^ 

37 



® ©0©®® ®® ©® 

FLgure^.l: Tlnj-jlHitLWntaiuJiilfiofticPS tor ("JJ. 

par^n? t-^r H^oj^cerns Lin: di r^C ticpn rir^TunJiiiialionf one i]citLi'. M tH A5 I [n? combiner ard 
the other acts a^ Ihc nimhtJLhi^ ThodLrGrLaon oE 4 cuiLbkitalH^n ta 1 rvarLan L JcrcfiB Che 
Lan^Bfe^ for W^rlpiri^ lIlc dirEM:tiu[i lit frnill riKht tc leEl- 

[3| C^mblQ^ two adjacenl Dctdfife. 

The otlLGrtbrte ja&rametejs of v^rIiOhI di^|pfncl OH t|ie cate^iy oF Ihe cOrrllmcr, 
»tid therpfciie the piiumpttr fl^UlngS are Btored lq the fcl^H^OJi Oil R per riitpgoE^ 
boEifl. Thfl wconrt paratneler cMceriLs Uie ^'^t^orial HiBLriciionfi of coinbi Il^^M^M . 
That is, some caCegOti^^ Jii^v coinbine with som^ faLcjori** and not with otherB. 
Tlie third paraiDeter coders Uni aiHlgltinf^Ji L of cateRory to thj? ri>tHt of tiLC tlflwli" 
ciealcd sl^racljl^. The chmceia reBtrLctfld t& eithw the category ol the left node or 
Ihe right node. The lAsi pirtMneierof varialior diclales ih^ |l^onolo^C4L level (i_tt 
T^ordn [fcllJ^tT or Berlence) at whirli Ibr ^orEl M irttion tales place. 

W^^tail now actouDL for ihree tif lh<- WaMpiFfc phenomena listed aboV^ HOhLkilLl 
and ^'erUal J^OmEfciKltion i JnhJ iiont] duoiuj ca^e pliiti*t^. Th^ PS fni decLnEd nouns is 
d^mon^lmled with thciioiinH tun^^i-tN. I'tiere are ^Vf-o elemental ]li>dfts it! thi^ PS for 
thiB word, one Fttr tli^ rcmn, aJid one for Ihc tasi^-rtLarLnii: Th^' rule of comhinition 
may *ppl>' hecauae thecornhiin^-ittjnJtJT E|tR tbeparametErh KtttiTi gs of tli i^ conihi nflfi 
tu. FiTBt of alU tlie r^oinhioc* is io the combinrr'a lefc, Jm inquired hy the languagc- 
WLile dLi«<'lion paracnjzter, Si^fi^ondly> ht^ Ijeing a. case-nutrln^. may JTomhloe with 
nokina at tlio ^tutl k-v^l kindly, we Bee that Hit new roftt is given the caEe-mjU-hcj 
fatejory- The resulting PS JS ehtjwn in figure 2 5. 




Figure 2-r^: Tiie PS for kur^u-h^ 

Nnnna markcJ For ^iv:oliit|ve taae are covered with ^ ^pwla] role of PS. In tht 
OVWit of a Tion-ovErtl> marked hiHtJ^ Pii fiit-ppl ie& a. nu 11 cateEory, tfiO ahsc^hl tlve cabe- 
m^ker. Tlda Oitfftcry Jim^t be posil^d bec^nf^' thi^ hliflolutive cafie, like the other 



ByiilacUi^ taiifiE, marks Kh^ flfJilH to which ll Lit irMltllLLc, and llin prwedlni^ notlltfl In 
thp phonolus^al phTAfle (b« hdo^")- "''t-* I*^ ^^^ the iJiBoLiillvc tirftUiLUfDJt of Uifl 
sample sentence J kurh^ J* depicted in Ilglire3.6- 




Tbt P5 for LEiEecLed v^ts U aIho acconfitnid fnif by the p.Lrtinif^terired rule. The 
dlffeience hero fOTifernt the CHMliltl^r.whicli at lilt ti^iise element, [t COmtine& wilh 
verb Ettms rither Ihaa noun*^ ard the r*yiij1tinE catefjory of the ccnabinatldtl W 
verb instead of t^fl^e elernent. Thn^ (ilflereiice Ln Inm^imlEBioD ol cat^ftnry t9 duo 
to tynl^tK effects, described hflow TIlp PS tor ths ve^b^ punla-rfit, is ^-."cn ill 
lifiure y.7- 




E'lgnreJ.ir Thfl PS fr^F punCn-mi 

The last phenoiflWiOlL hictdled by l.hc ruk of combinatLon In that of contlmoua 
case phratts- <""cnt&Lder ihe phr^jw In (^)_ In Ibla phras-^^ the cAEi:-cnAfker, -fVh, hlft 
flCOpf o^eraJ! IhnK nouna This anal>lfW 13 sraminalical l>iwaitfie caEC-cawtHirB may 
alsu be adjacent to n^iUti^ at the phraM Icvnl (In additioii \<y th^ word level, Aft ^"Uh 
the declinoJ nonn^ abo\-G). Tha PS for Ais phrase t« ihfjwn in fifiurt^ 2.6- 

(j|J ^irrinjt vf^™^ tanJiFTpn-rfu 
ccntjpiu^e kiomesLck hnLve-i.B^ 
'the bnive^ hoilie^Kt cenlipcdc' 

Tlitrt are two well-rormH:ilnWH rordllionb for PS- The Hizt, £Wi?li In (5)^ strttf^ 
thai PS niuM not HTOIiLAJH Jiiy iLTiCDi[Ll]inrd ttriictarp^. Hn^'i^ver, tliln JTOliilitiDn b 
tm i^vere for all lan^naEn^. Iji W^iilpiri, fi>r inhtance^ thet^? Sire (lo arderinfi cnil- 
^LTaiikls trel^'^-^^ci phrafi^T and so PS Will not tontain ^^nnnef U>d slranliires betwi^en 




yirrcnj] 



firrftTl^ 



itiTP^aj-rpa- 



FiEDrc^a: Thi2 PS for {4j. 



phT^SES. Tbe facts arc LOVGied with a pajameter of v^n :it in^ji lEihI fllctat^ at v^tlch 
phonf^lo^caL levels th** ronflltlof ^|Hpli^, TIlLJ COilcLltloal appJlM botb at tliG ^'ord 
^nd lik^ plifiu^^ li?^'^?! lor WrtrlptnL linl not at L^jg BeDLeoEie lovol '1'h^t.R^ the P& for 
"voids and phraseg mutt coQtaJ a £] Dgle atrkj^iturefl^ biqt ther^ J$ ii0 4iJ']ll r'^uircnunil 

frt the RepLPitM l?iH"el- 



[5) PS iniiat he fuHy rorrRrtrt]. 



Thifl 0i [Igjt <Oi]d] lichti J iL f Onj ij DC M I7D wUh llii?pajajnetei5 of variation for the rule 
ofccimbiDatioiii-Rervefl tonile(Mjt mimy typ^ ol iingram inAtkia! WOfvlfin (nLrasEi^, and 
tentfn^;^ that a.ro i4Jjgram]n^Lki^::i] with JnSptcl to prH?tJ]dH^ntfl. {(!} coniaLnB three 
types of nnj^aiuioatkaL T^nts, all of wbiirh are rul^^ out by the weH-fotcnn^nofi 
roTidktiaih- {a| £lveA ui exani])le of a nj^t^^r^* uubU]4Lt[h7 tenEC dementB may not 
combine T^ith nouns. |b] Eho?^ a woid with mi^rphen^efi m the ^rong Cnrdcr b^ 
ran?? verh fftemg do not H:o]]]l>mo wllIi teaiar floicientsj iht l^'o cfem^nl^ Btnicturm 
won't be combined antDon?*, aj^d t|i^ WJ Jl -f^tTilU^I tjSJJ i^MrfitJod Will ru|ji tlhiG Word 
out. (c) 4h<HW» the verb with thr rn^mpcnc-uL^ in the nghi Drder^ but in sepaiaLe 
^ojdt. BecauBO the tense ci€:m€nt iB con&LTaLni?d to fDcnbcne i^t the word l^vel, tid 
combination haJ^tB plare^ B.nd the nonditir^n rii|A? thJ^ one out t04r 



(fl) a_ ^tnrdu-mi 

tMld-HPNPA5T 

Cr ^ptrrpfd TtiJ 

take NOriPiUT 



ThsEipcoiid well-forniediseflacQDdnioti rnnr^rn? th<> fOJrt fKflUloiL of wj^tds icL Witl- 
piii, md |fl HtJ4.t^ Ml {71. Wofkls tlL^^t H_-J3ilFiLii jusl one syllabk aie not gramraAlical, 
LfLdt^^Jul llhri^y musl b^ Eiiclhtic to a precfdLn^ wopd. Tti^ au:<Ma:&ryi i^, la ^rt ^X^mi^lc 
^r Huch ^ cLllJc- NcpIh? tti^l li^J lLi>UtLS linri v^rbs anlOmaricE^ly pasE Ihit condition^ 
Nut fii^r dilTerent rea^oii^. r^DLJi]£ pa^E b-HauEO, afi "le^mi: Il> hf Lh^ n^iiJ^r Xh^.t^ Are 
in? si ngli^'gyLE^bli:: ili^ljil Htrou^.'^ Verbs pnifiii because Lh^^' riueL beinA^ct^d for t^D^Bq 
there are no null verba nor tenK? ^km^ntJ, 

(7) Wordfi must fon^iAt of al [nn^t two fchvllableu. 

2.1.1 Auxiliary Compn'^Hioii 

Th^ fl-ijxiliary 1^ ?til irr^pjLLT wniirJ. Unftirt-Tj iLiLel^j only a descriptive theory of iLfi 
compogit-ion 1h avwI^Uf- U conaJsL^ olft nkLuiber of t[ii>rphc:mc!^, ^U of ^vfLirlL may i>r 
m^y iLOt be prOE^[it in the 5i]T(acestris^L^ The pa^Eer covers Lhe part of the auxiliary 
wciricon&LStltigof theliajtT ihoaaieemejttclLllcHficid tliodilive n^^itratio]! [UiJ-k-^t". 
Theif positacmiuE witEiEQ the word Is best given by a template, as shown in li^Liie^LDx^ 




Fipare &.fl: The auxiliary template. 

The PS for tlie auxiliary la ajgo bmlt sip wltli tlie rule of cc^mbliifttlon, however^ 
two m^liliCfttioiLfc: 4ri? teqjir^. FET^t, l!he categorial metrktions cf the combLaBn 
mu^t :^]oiv [n^r th^ epiicmali ty of the elemeDts. 1'bat ]e^ tlie combiners may not 
h^vG static categorial res Lri c tLons: insLh^ad, tbf reHtrkctions must be dyciamL^ly 
det^maned. di?pern1|rij on tliC ovisrt naOtfphtm^ Sequr^mri^. Tilt ti^mjjl^lc, iJjuVnjs eh 
cQUfinlted to ct^t^mlnc ftHUdm4tic?il a^u^ntts. 

Th€ second chanj^e concemt the couHLrurijon of Lbe auxiliary structu re. Ralher 
than conibinine ftdj^ent nodw itito a IjinAry trWn th^ iuxiljury morpln^mfg are 
combined as s]blin|^, chiLdTen of a aingle parerit !>ode. Thi^ hnear structure Iff 
TiEcd to rfidwt the Himpli^ trjmpktp^ thit dtscfihti: [\\<^ pos*ii^le cciinbm&ttori of the 
compiui'^nt murjiheme^. A:^ an e^anlp^e^ the PS for ta-mn-r^fr, Lhf" au^llary o[{2}, 
]s ^v€n in lij^re 2.]f]/ Note thaL lIeir auxlh^iy "^'ord cc^ntaJUE three of the four 
pos^llilc mOfphctriH^^, omitting th^ object, agieement clitic. 



Of raian^, ika u iZDCjrcLj^ Ln rmjuricRl jwjaL if it IhSTriFi finl tJidt uo^^xfLiAbLc dolieu exh^ 
Lh?y Lijij wi>ii]d \\ii HLihj?cL Ld eIilc cocidLtiDa 

Srii [NMitiCl fiiii a DLDr? dcr^«d Jlbchiiigd gI ific ^iiKibary c deh ^imicdI*. 

cvtr, tlinr F^^jir^ rr^ii r? id tbi' ^nxLlt^rv ^^hJ i^n )w n^-i l>ii til^iL f^tr 3h^ ox'H^ndivii^ ili^ iH^nipkii^ 

1 be rzitrjiLry damlniLLn^ l|in Akfhiil^Ptry nvbril ia Kliplri] ^A^ t>T " AI^^-'' A1JX ^% ^d\xc^I\l\J\\\a\\^ 

lhillkr'>LiL'i I rh W^rl^lrii wIfIi Ihr I^'elh* ^iid j^r^yniciit LiiEiirmAlKJ^i M^nuL dvi r LliG li^XkJit^^V ^lti| 



Al 




Figure 2-10^ The PS for ka-fia^Tla. 

2.1,2 Auxiliary PofiUioiiiiiB 

Tl,^ ntfiiiionlns ^r ilif ^.i^xiliary i- qms. Jiilt* rh*t nf cih.T words. R„iishlyspc;^Ji- 

Tlilfl flped^l prnperlv is ^ccailod for m ...n'^ij^. fli^l, ^i.SlllTiry ^orf. (wiLl, ^^ 
™pno,.H U- h-- d^fTibpd hlh^> d^ mi .mriblllH^ will, yLher ^tds, Tlial iH the 
,.CH^fP« ElnKru.eE [or :,i.^iliiti.3 «c? lnH.'rt. Sct^.ci, a^,^:iLlJr^^ ate cOilftid^-^d 

fODEliainl is beil ftlaled -Jlh respctl Lo ite pLwe in [l,n Sf"lQi«? ^ a wKoJt Tlua 
weLI-fosulrJilTieEfi t*i||(IIIJDQ it ftlven ir (0). 

(a) The aiLdlkjy muBl flpp«Lr in riLher tht (Lt^I or Hfford po^iti™ 

Fhw «ainpl*. CDDEid^r t!i€ p^fllric-omp of the anfJlliry *oi(l Ir the PS foF the 

BEDipIf seolctf , 13). Eho^i'l. In fiEi.c- 2.11. A^ rn-tiliored above, rl,r; fir.l Lwo*nrdB 
urthi£ EtnrfUCPir^-^nUioe-l in ^aii^le plPr:i«. Drf^i-ii^ iUXillAr*. do ilOt combic. 
*il.h0lhPt*OKls,lliePSforlbcphnLaPCDDEimHjri*otlriirliirG,,UPPforFis«ji^^-i'rJif 

flnrf me rt^r la-TH^-rr^ Afi lui^iliarira ^.f (^xempL tmn. Ll,r rur ii« led n-fls cord,li"r, 
iMa phra^ is nol ^o.iELden^.1 ll l-fcnnH^J . Ttifl ptfi.licimne l" ^h^kod, How-er, l.y the 
wraUaiy p^ltionire MildiLion, which llll^ PS p«tfs. ^ ll,^ lurili.ury wo,d ,b Ih-! 
tecDnd BlrncLuie, 








1 iRurpMI: rhnPa rbr(2). 

tion,e..,SlLflria ...l.»l, hi r«l, c.^mbini? ^Uholhet word^. Th- «repliOnh COr^iEl 
dT Ihc anuliariffS (hal b-.-^.n willi .l rllH.' moiiiliPme: Ihn-y im^l b<' enrlit,c II, a 
],f«-.diD!. .^r.l. ThH a.ivili.icv bJi^e, Vi. ^ "-" « "i" W«'J"ei>t °'^^'i'f& ""^ 



A? 



tbe i!a-tlvc rtgUtration mEiker, ftie a[l f:lilii:B. AuAiliac^ f'litn^iajitloii iJIFEcjis fiom 
i^Urmil iJiiiciiitioii {e.g., a* witti case-marken J iii ti^-o rft^pe-^U. Firatn auxilLariHRs 
do Hot have Jiny fatcjoriil rM^iciioiLs on ihar r^miWtlCHi. Sec^md, a£ with rDi]-r||L[c 
AUXJliirits. no combination flf Structure UikL-B place beta jbc h: L tic ^HKiliiftriF^ a^E also 
liivifiible to the f^tinrMttdjUft:* Coiidilioii- 

A grammatical example of the \^^ of ji clitic iUxjljajy ]e jpven in (9j TIiIn 
sen^Ence is jiq^t likf the eiimt^L^ E^clEncE^ EJxepL tlhaL |t i^ Ucklhg ilhc imp Erf^c tkve 
haae, ia_ EewUM the auxiliai^ word hegiaii wUfi ftn agrn-^jmcnt marksr. it must be 
enclitic to iks prfteding word, ivlikll il it. The P^ Joi thl-5 spntence la ft^OWI] in 

(9) ffffttjidu-riu-f^a-rfG pufita-mi tuTiiU'kM k^rlu 

l-EfLG-ls-;id Catfl-r^OHPA^T thild-DAT boomEiaup; 
'I Will taks tb« bocmierang fioni lK^ rklld." 

Btfoay^ linisliiiifi the dLBCiiBsion of auxiliary pOBltLotiltign iconple ofpointB &limjld 
bs mentioned- Hr^t. uolf tlutt tbc reqairement tlip r]ltir, fnisiltarlcfl rtnifit Appear 
in the Second position foLlowp fn>m tins fL-Hjajrenienl that tbe^e words be enrlULC 
tn & wordn ind tli^tL aaxilLarieE most appeftj ir fflttier thf /irat or second pofiition 
(following Ihs well- formed ntflfti:i>ndJtcCsJi. aboi-e). ThepoBitioningof slii^rt fl.i3xlll:^rl« 
[irf-T tEnMecoiliitttinEol a hin^ssy Liable) is alflo :tcccHL[ltc:d [i>t- ln_-rj?. By the condition 
stated abov^ ^Udla auxLlilttics miLbt be enclitic^ and ao th^LF pCKhtloniilg Ls ^iitllL^rly 








Figure MMV PS for (91. 



2.2 Syntax 

This ffecLLon dfiBCriL>es the &jnticLic component rnf CH Tlieie are three kvcb of rj_'p- 
resentatlfin within thiB component, |_ n Li urt tj rt;, D-attui turt^n -^nd S-BlrnctnrJi. Both 
L-^tructnre and D-itruH:L*if<> rcpJu^tnt IhesynlartLC manifetLatLon of p red i oat ive re- 
iTLlion*. L-EtrnctLirE [s connemul With the predicative natiue of individual IexIcaI 
ilsmEj wLtle n fttrntturf contains thE relatton^ *tf tJie aontenf'e m * wht^fc In fact, 
it construi^tttT ^ 4i combinatioD ol fon_Ht|LiLent Tj^^triKlicMS- S ^ttuclurr- rjn ikc 
-Jthec liandj is conrerm^ jn pirl wirJt cnLwr-marking relatji>n& betwEEn caAe-]jiarken^ 
and th^lr ^rgitinctitf. But thsse two BjntMtit Views a-re i»ot wtltogondr tlitn? ii 

41 



An im]icirLajiL idea bwhinil th^i- lljiRuLRtlc ^Iructureid \^ tin? fLOtJnjn <vi Hira^stug- 
Tlie atruciiirffi coiit.im rfilf*'Tont el<'mej!tji, Jiomc of 'l^'^lll:^ allow for t||c p:<iHlf^Jice oF 
[il1h4>T« by Di:*Jlsins them. A prinni JiXiiltlph of lirenfiinj; codcctub Lfn? i^nzJk^tur 
and iLs ar^UinciLl* The pr^iBonce of eh argQinniH \t\ ti g^nt^rme i& due euldy to 
ttc prMHTiitor- At D-stniclurt llkis llfflwlT^ft concercfl the MBL^inct]t of Snimaiitlc 
roleo ti> argumentii- onlv those roles thai are |^rl dE tbe predicator't meaning are 
Ik-iLied in the atraLture. At S-itrui^turtf "the llcenaing is for caBe-markin^. Ak V-'ith 
D-stxactures Only tho^ ajgunientc Belected by the urcdifator Aif- lii^^nnefl loj nafle. 

CDiiiidtrll]Ji5Jim|i|e3eaLeiiceT [2]. Th*.-]]roJltatorl?. t|ie verhgtepp, punta, whichi 
duo to at* nieanLng, licEaBet lhrfi?c sttgUltHItltR, the Ifttefs the talcca, and Iht Eouira 
from which Ihu takn^ IH UtV^H Purine Ijcooibm tlirM pOSltLOJi* In D-aiTftCCiire for ita 
ttrgumtnta, h S-Ktructura^ the I'eiln lictllSt^ llir-^e J^JJe phraaes (lq tiih mstanrir. 
ITtarked forerg^tlvt^ ftl)^i|t|Vi>, ajid dative cauej. Th<A^ l[^%ViS<^ ftrgum^ntA appsar 
ih tht senlonc^ aE ttree caBC-marL^d nOufb^n 

All three a^Titactic levels hn'e the flam? bafiir Torm aatd cortair the Bairte btek 
OUtlti^Bi F.(!., syntactic' fategoiiefi Th«r BLxacturtis givn^d byX-theoryn rfeHrrihed in 
lie follQwing Bection. FoUcrt^'ing thai <liflC UflSlun ttjmefi * descripLion of eath of the 
siqbl^vcla and the luappilL^ betwi^Q th«m^ 

2*2-1 X-llitiPl:^ 

X-lheor> givetj Iht 5tf ^ACtiiT-^ of the syntactic reprsHcnULliJiitH ftfCB- Tin* fflftln Irt*** 
hehmd this tfiear^- tb tfui-L r.itlk basic Itflm [f.y-T nouQ or verb) ]& the ceiatral tlnirtn^iit 
of Ltfl own |ibriise, and that the structure of a iCCIteflCf ^ron^let^ of a rombination of 
these fitructiirea. The central JiLecnenl* areH:alled li^ds, and the BtfucLures of ivhidi 
thty lorttl the n^n^e ar? called prcjccUuns. Thf ltF^M! (prfljei^(* pnme number of levels 
to fcnrm the projectiDo. Th^ tii^fifit Itvet o[ iJkf prtjjection is called tb^ mczima/ 
projection- 

A major t[^]iX\ of Xth^ory is that t1]€ same stnictcLte frn^h^inii :i|ap|l4^ to all 
tatCftfirlw; a|] phrases [c.^., ooan i^hnu^ aiid vorh phri'ifls) are afiaumed to haTe 
roughly thi: siilHi Strut turt The number of leveh io the projection 14 parfUHetertp*d 
at] a per cat^j^orv baaisj lio^cvOr. L^y^H^al Iti^mfln 5Uiih aa oountand verba, projucl two 
levfla;^ other IteoiB. auch aa cate- anlrktt^ ^ pl^jjfMit ore The aiixiliFU^ pcojccit tv--o 
leVclfcpaJl tSJJlained below. Ftjc example, the Y-strai^ture forverh phra-^cffl ia depinted 
in Gj^ure 2.13. Aa ivith Vii, the syntactic EtractureB of Xlh^r^ry &fi^ dej"ttrtL wltii 
Bodefi connected by links; (ht l^v^^ ^ij projectioci is indicaLevI with tliC number of 
bai5 abovt tht tat^jt^nal label- 
Tin? purpose oftte projections is to create slot* in the i?t rtji^tvrefl for tlii2 attach- 
ment of otitfr projCCtliMS- Thefle alots art manifest ^ siblings of the nOD- maiti cfjal 
productions- The siblins? oPlhe hejJ arr^ t^llerf tirnnpir^nt^nSf^ ajid the sibling of the 
Urtt-lovclproJtTtlon {ftir two-level projec tioas) is calhsrf the ^p^ijiEr, T li*ae si hlinj^p 



44 



Fleurr ?.I3: The proji--i:Ui^ lur a verb pLrase. 

Iti f^^rfJ AT« C^LIecl ilry^rtlcrU*. vT- theory rnrther atatsE that ChFJ^rgEiment^ofaprCH 

jectioEi inuat Lhemg^vee fc>f maximal iJtOJCclJoiiB. Th]s is diaj^rammed m Pgur^ 3-14 
with a IWo-Ie]vi±I prGjecLion. 




Fi^re 2*IA^ A twi>level projccUoci and its afgumEuU. 

Bui LLere h ^ prohlftm with tin? tfSdUl&nal Iheory. ^-lti€cry statE^ that JT- 
at runt urea ^fi\ omIothjJ hy pTttedemcfi, in adHlLii^n to li]trar^|iy_ H^yw ^slu weftCfJ>unt 
for Hit frw ^™rd orderinft iJ" l[|P ftyntftCtJC 6tnirtiirt¥. ^tn: urdi>r4.Hj b> prCTwfiratcT 
^□me pf^rf^tly gr&nii3iatica[ 5GI1 Lgciccc T^auld be iiEfiignfd |||-|4^rm^ HLri^^tur^^H Ei liv- 
ing tfi*sins arcB (c_j., when a nai^l pk mr tl t prt^c^jdw thn^ spKifiiir). Tlni vtri-ioiL of 
A-thfory prf-SfJltOnl hi^reU nnl ordered h> sLblfcn^ praceJence rntteadjCmly thedom- 
inanr^ i4?JatiO]is are r^'pTecented, u'bat ordering LK^r^ lk aniGjig ^ut'Ju:^ izoT^aLltu^nti 
1b njpreseiileJ in PS- In lin<? wltlL Occacrt'^ razor, h>iiiacl]c aLructuro need not rep- 
resent pr€M:«li?in:e hrfftltst Enre^cdeDCG must be contairod Ln PS Lo account lor Lire 
hnearlLy Ipbi^rc^nt kjL lis level ci reprDSinntalloii. Tb<^ifc|^ti Llh^ deplnzLion^ oFfi^JktacllL 
i^lrUH^tiFromnst bosbciwn rl:*LterL«l OJi W p^igiSWilb adtrccticuiHtmong Hie coi]Eli tnEint 



4A 



l1i'rLpri> ^'nLcTkJk^ ^dhHcutsion of Lhf ay iita£ Ik Li?vds Lfii^mficl^^sUi lliiif ^c^tichn pra- 

ofY-lheory, Rovern m i]n I, ih rltlim^. Tb^ ^Wtlrto conrkfLfi? with t be at muturfl] defi- 
TUt](3ll cf tbe graminaLLiia] fu nctiona^ aubjc^rt rhtin] obj^t^. 

d^veiEtniGnt Theory 

Tlir; (/^t^ettifW^il relation \^^i^ h^^n loiipd to bo ii^Eifu] in cxplaLniiig many Eyniactic 

[Uhoei]): 

C^-4?jgtyatT?4Y^4'.' A r -CQm FTA nfia Q If Fii]d only if Uie EljiL branclunj; node 
doniLJiaLinE A al^o dr^mlclit^ B^ ^LliJ A rt<rffl not H5o|f dnmlmti? B_ [p. 

For examples cf c-Eoininaiid ^ cciiBader tij^ur? "J.] 4 ^tg^ti. TlhC hCad C-COi[kmaiidJi 
Bach of its cQmp[E>]]hi>nlb, ^hJ, In E^N^1.r tli^' c[>[tiEL|o4nchls c-comitiand the hc^. Thfs 

tbct^leTEl prajectjcn r-CGiunjajidB Lbe spociFi&r, iind uic€ cvra^ 

C-cDinmand, in Liinii u ue^ Lo cIgHilp £flVOrtlini?iLt (also Uktn from [vRW36]): 

Gottf^jntfttt \ gciverria Y f1 4nd only if V l^ CH^Dtoined in the tuaxi- 

malK-prDJMtLcm crfX. X'"^; X^" \s Lhe Binitllisl tn,iXtliLil lirojcfllod 
i^ontainii^g V: spd X c-comniandB V. [p. 1^1] 

nly Tif>Ti-1'naM liiaJ pToj^LiDn^ may ac L aa govErnciE [^X^in thEdcrLnilionabcVe].^ 
Afiaii] reJerriDfi Lo Tafiure ".^,14, wt^ set th^t util> Hif tpftlliet' :ifl<! fom plfltllCcl 19 are 
govcrcied, and that thojr ^ole governara oje thf Eijat-level projecljoiL and lbe bead^ 
twpectivdy- 

G^AiHtn^tical Ftiiictldil« 

Grammacic^tl fu nt tions ixt^ dffinwl h twmp of IbeJr XposhJoiwi:^** 

5iiiyifrlr tJi^sihliEig i>f th^ one-kve] pro^KLion- K, t £_, the BpEciHer 
Object: a Bibliug of lli^i wto-lfvjsl projcrlldfL, X^ i-^n ^ complcmeiH 

Whale tliprc nLiy bJ! suy nuniboj of ohj^rta, a£ dictated by oLhEi aapecta of 
p^r^tiiniar, Lhsrs may only be a ain^e EubjErl. Iti Pa^t, prcdkatdr^ ^4 rE^CiuEnid 10 
bavr ^ HLihli'i'l, H.s ^-^lained by tJ-xtended PToJection PrincLplE (taken from [CboSQj]: 



*Th^ ilhrfiri^ -^IJlIiUy fr^n ^ I1i=i^ inland lti^ BCitJr^n 4>f pr^prr ^owfjmaw, tkIlicIi may Duty be Lciuti. 

WLLb X-Ll]«rY ar^niDCiit ponlJaDH UcrtftVct^ iL-fi i^fahjl^4ti 4f duI^jH:! for IP^nnlL^^ Ihha EDfllpFh 
ilfiM difftEn l« IWiL44fi fOE nLUnfi*! T^ tp^i^Icliti grb^r.iliEj Ifac ^u^LL^T1 ^ gramEnLiLc^ f[3DcL»Ti Ji 

M 



Ettfiiidp^ ProjfcU^n PriurrpI*- .,. Ih^ n?quJtJ_^mciLi i.liil cUusirt have 
AtibjOLia ... [p. EI6J 

Tht hociotia ofKiiijjijcl and objott iro cnotlviicd byccHntrol plL^notdniUiu Control 
theory AitftmptFitn explain ihf m tf Ti^reLH-t i on of t h^* iTripll<^l L iiTp;upifliit chat i£ pres*Tit 
Ia iufLcLlliv^ sub^nJLEi^t^ d^us^. Thl^ pli^in^cn^noh Is bn^^ ir\lv\y<imc(^ by way of 
an «x:uiiplo- (lOl hdtiw (frcptn [HalS3]) ^bowg ihvo eKa-inpl** d auhiMt contrnL 
In instuicefl cf BubjecL control Ibe [inpEicit Hubjecr of rlbFz ftahordiD^itc clidSG Ia 
HildcrfttcXfcCl Lta be ^orof^JT^[lL wlt^ tlii5 -^^hjeft L-][tlii? malri>r rUiiae. The prfsenc? of 
Ibe subJH t-ccmi ro] compl^m^nliz^rH iorm p a UeU b^fd ic Ml? Gnibcildod vi^rbi indlnj^^lLl 
LbaL thEse GeiLteQce& jj~o LnBbuiceE of Bubject rooLroL It is appropriate to refer to 
the notioti c^E gT^mmatkal fuDctior, ti^th-er tli^tia ca^, b«caiji^ both i^ubject <^j\^ 
(ei^tLve ani] ^bfinilutLve} appear ld tbLa Eoostructioii. 

{10) a. /"^ffftTTttt-npil^ itft pijrtipfl Junpfl^rtiP iarJi jorr^Ii-rninja-taminrJN- 
niAn-Khf^ (HPPH-F corrohtjree Birg-]J^>KF'ASr boomeraiig 
lTlm-[?J F-COW P-E EtG 

'Tln! ut^h Ifl fJn^ing a ctarrflhoRe song wliile iTlmn^iTig the boesmeTang.' 

b. Aam^a ka-j\i i^an^a-Fni ^r^o karSa-fYJa-karTUM 

Wonaan IMPEJtF-li> 6 pj?iik- H <^ N FAST yam dlg-IJ^r-t'ciMf* 
'The WQF^FLTi Ih Bp^&|c3T^g tc iHE wbile d]^[^Dj^ yanu/ 

l^^r^l le]]i n £ th? ^uhj^L coiitTiJ e^ampleB are exampleE cf objecL rooirol (^ko 
from [Ha183]}, gvcm in (Ll)r OhJKt HTHMitn)! Is llit [HtLCilomenfi^n wlipre thf object of 
the TTii^trk^r clmji^p, nnt Lh^ ?iih}^t, ]^ imdeTf^Lnod to be cnrpfereriC with tbe Eabjccl 
dI the Eubardinate clause. The object-con Lrol coiEipLeme]b LJ l^er^ ^SO att^hed to 
ttit ouiheddtd verbfc i* i«TTdr AgAMi, ]t J* ]aro]^tr to t^nploy pftflim^H-tln^al funciioiia 
instead of case because both objecLivc casea {a-bfolutLve and d^itjvc'] ^vr^' used L]t tTJs 
type o[ sejittoce. 

(11} Ql Punfa-n^-ns'j l:a^f^a-n^(u iran^ta - njn - 'jEimi. 

ftumi- pertei ve^ h M f aet I m ^^ t n f - Is^Eo h ppak-f M v-co M P 
4 hear ycrn speaking/ 
b. .^'yflrvta-pftCu Atl-nw-Jrtnfl »ja-n^ (jMtuim" panli-mifya-lufTO- 

man-PAUCAL ] U PETIF- 1^-3330 &ee-UDVPASr lian&troti ApC^r-tU^'-^c^MF 
^I flee Ihi? &evend uHjr Kpflridg th^ t^ialgbJOO-^ 

kangarOfJ-UAT IM^EEtF lfl-3d ftaalk-wOr^PA_^T jrafis eal-liar-COMP-DAT 
^I am fincakin^ up on 1lje k^uig^mo (i^'llile It Ifl} e^tmg gr^js/ 

Tte control relation is bEsiin3«i to be & Etmci^r^ Otie, and th^r^nr^ tKe Bub- 

ject/objecl aa^mmetry decnooslialed in Ihe exajDpleff abo^'c ]iiual b*i represented in 
a iStrUHZtiirrLl rtian]Lur. BrLcd^, this is irfhiewd b> pla<^ina th^ Bubject arj^ument at d. 
hi^er level in tbe projeclion tlhan LheobjCTit: th^subjctf bitfSH^ttated iailK Llieipec- 
i^er pciB]tLon^ and like obj^^ts are asaocLH^tod wUb the ' cm pis: nK-n t po^ULonf:- The 
cheilitis Chf ^Ihbord mate ron^Hl^nienti^^iT either iufrnor iLrrd, then dictate wber^ the 

47 



EDbQnlinftlr plir^^RinrlcGr -should b^ nlUirli«] L4J tli? m4i-1rL>L ]?lir^^^' }i]7ir]i[T. I]l LEic 
ev^nL ofBubJoct colitti^], llK' ^ ij horiJ Iji^lf' pLra^c lb aLtaf h^d bo thzt iL e^-ecii] m uiJe 
ILk' aubjent^ tar object EnnLrd^ \bc h u burd iniiiG ^hra^t r^coniitiatinje i]\t i>bj^t. 

2,2,2 L-strnc-lurfi 

L^gtrui=llir« represent the B^nUciic ]uaiiLrDiJ titt Lo}^ c^ricjur-^ It^iue, E^h Ll^m pro- 
j^U into a ^irt^f^ ti-atuEic tniG^ which is a BlDf;le A-projcc Lif^ . Aa mcntiomzd FibavG^ 
l.lic Dumber orlflvdn of [^rir^jftr L|ob Is dPtoritiLiied b> tho ilem'B catosory. Hci^'ever+ 
t^i^fe III fnoK t hat d^Cernii n^]e tb^ partacuSar in^aLiI-iz^labkQn uE difTirr^iit l4Vj£n[ iL^^tris. 
GB 4:lAiui4 tll^t p&rt o? ih^ Boni^ntiE: coiitsnL of a FeiucoI Item \3 involvGd Ln ihe 
(En] vatLC^CL olltf: iiyntArlic maniEEH Lotion. In piirticiilar, t]h4> nuinbH?^ itnd type cf jt^ 
arpimpiilE dictitt^ bow thi-y wiH ,^ppi»aj. The tlrflt p^L cJ ihl-^ srr.tlnii sliscTtsftiw 
fl-tbi:oTjj "frlucli attempts to esplair (h? mapping fiom ji^^niJiTitir tn ayntartic fl-rgu- 
nient?, J'h^ other put do&cribcB tl]o BaiapL^ a4?t qF rul^ LJkat LiidkaLes wlier€ thi3 
aynLiu'tjc ^r^umenlB aT€ plar^ in Lb? L-?lruclun'. 

ff-thc<ity 

The mi?ai]][L^ or toxjiz^ Itein^ mui^L i^onl^in ^ a jninimiLm infEnrni^tion Kh<mi th« 
ri>niher &nd typp ot mgum^ntF^ tlhat it tak^a. CcniBidcr tbi? ^erb from che sample 
acjil^RHX-p jrtiFilo, Pari of wbit on*^ trowg :tLk>iiL tfl-kirg t^ ihii t-li^r*' uh i tak^j-r * 
Lhii]g which ia latins and hl aourte froin v/hirh ilhC taEiien ia taki^tir Of Hioune^ tlicn 

i^ mor« m^i^Jilfig^ but ttiLs niu^Ji ^esnu rnlQiniilJ} nscsE^ry. 

fl-lbeory^' iif cHMiccm-id -A-itb iftplurhig tlir [LiLun^ of iliMi^ **ttiant|c: atgtiini^ht^ 
and bow th^y appear in Eyn^ix. ArgnnienLf arc cill^ #-rcffj (tb^criiLLk roLos). 
Altlhougb tlie tb^ory of V-roLea 1e &ti1l qiute fuzzy^ a lev^ lotes seein La LTop up 
fepeAtc<lly, Th^ moat common of tbfse an? AGRHT, Ih^ pprTornii^r oE art EbctiojiT 
TJ1CUF ibe abject aJTected by an action, and PATH, tbi^ Eonrce ar go^ of an lution. 

A predicator is ^aid Id ScS^rl a numt^Gr of ^-Tolm. Tb.^ Usl of d-ralc£ that a 
pr^lcaCc^r geje^ts U ratt^il a 0-ff%'id F^tr example^ lIle prgdEcaLor^ punta^ b^a a 
0-^id that contains Lhree ^-rales^ .^GEMT^ TElEKfE^ and PATIT. HuwevEr^ doL aJl 
tomhlJutliMafr of fl-rola ocfiir In Wirlplri. Tlie fl-grtdfl that do appear are Hated In 
table 2.2. 

E^-ro^ES appear ±e uynUctic r^U^gork^s tli^t arc tnid to b^^r tb^^ cOrreiipon d I n g 
rale. The Lh^H^ry u^ed berE a^nmce thai ad ar^niEiilB appciir aa cq^g p^rojf^.^^ 
Ctt^^ pliFw^ rrn^ivi: tlnslr fl Kflf unrfn^r t h*^ sy n cat r K rel^tton offl-a^itijrtpipni, i>r>drr 
thE rslatian of goi-Ern [ntdl . Tht position Lil -JJlikb Ibty re^j^ive ibn^r ff-rol^^ ts c^Jlcd 
a ^-po^HennL An exaiuplE of ^-aBEij^nmen t wLd be ^veh lh thE aEction below Tbe 
Ibt o( Warl|>lti fl-a^ftitncB fovi^red [yy like p^-rei^r is ftjveji m tal>li? 2.3. 

Canaidcr IhE vi^rb pun^nr once iif^in. FoLkrwiiL^ table ^.^, we see tbat botli tbe 
A4EINT and niEMR fl-rtjloa are^ayl^iwd hy t[\^ verb, "i'lir- verb ia rot JLhJe to M*iart 



'^S«i, for fttCTipJi^T |?Ui&l|. 

'*T*" t|^*w■ry m ipi pciri rtf.1iiril. ac H iocH fi(K r^cirrnhnt fr>r ^Ttiii iikii t^ nf vf^Iih ■■rbHH'r Furii^t^Jn*. 

be dcb^ncJ iar n^v 



F-?inJ EMHJl 



1. 


T1IEU^ 






yfl "111 gn' 


Uh 


TJIfiil^ 


PAlfcl 




^utkQ 'Id Iqvg' 


3. 


AQbETt 


PATH 




Tiitim" 'to E^Jtik^ 


t 


AaewT 


THEUE 




n^Q "la sbg' 


5. 


AGBrtT 


THEMLJ 


P-ATH 


;?^r>ffl 'tn tjt|«' 



Table 2-2: TliE RvE verb^ ff-gndfi, 



V AGENT 

V tHtKie 
|]AT PAT][ 

TalilE?.3: Warlpnri ff-flBsiETiert. 

all oCllie ^-rolet that it licensess hnW^vcr. The oif L^iiiDd Lng S^fc^k, fA'L'H, is asiigped 
by thp dttlve case-marlier. DAT_ In order Lo aHDC]iile this ^rgdlll^nt wjt^ t|ie verba] 
projeniLiDn- ihe verb mu£t j ncf trcctf^ uel^ the 0- role Ihrcugh IIlc d^Uv^t^^ marker. 
Imlarettfl-aMigilincaLt ils^Cntattfl piM* nnrter fiQVL^nmerit Tlkia wall be d EmcKistrit«l 
kn the EectidD on D-ElmctuTeH bc!nw. 

Placing Ih? Ar£uni«lild 

fl't hcory dltlales how JP-ro|i=a appf^ar Ryutaciticanvj and whltla cISmtBta liftilEt tln^lTl- 
T be r^-Jnaming qncsLLon tc answer Lii where I bci^ riFgujEii^ELt^ appn^ r Ln l^i? L-^tructureL 
This iil[H>rtn4t]on roii^^ from t\v^ ma-pplng nfS-roLes to Riaminalical liinction&. That 
|g, the mappLni^ apcciliee which arj^moot iipponiTt oa ilhC riUbjO^C. am] which ^p^- 

The mappiDg lo grammatkol Tnnc^oTi ie natdlhted hy the rtl^incticin bi^lween 

fxlems/atid pnierpjft/ft^roMWtlSl]- In the ftLtndajd form oi the theor^iH ihc tJit^r- 
nal tf-rde a& as^ij^T^ed itt role ou t^ide the iTLaxitn^l E>rtij^t]0[i oflhe pre^lcator, and 
iatcmil ef-rOl»i an] iascEcH^d thCLi rolj^i wilhltl tho mfl-^MTial projEcLiDn. UnT-nrtu- 
IiaLely, ihcTe ii^ iic Eolid theory erplainiii^ which ^-roleji are c\t4?nial ^Wld u'hlel^ ari^ 
inLercal. For Worlpiri there i^ i iiinpLc nile that dlctftt^g which fl-role of a fl-Jtid 
will he extemalj 

&ftrTj4iIpTf7ftr T[ tht AcjfcPT #-ro]e Ih ^elefted^ tlacu ]t ia the nixterii.fcl 
#-roie. olberwL^e the TUEME is 

Th^ th'^sry of ^xtrrn-'Ll antl fcuLernfrl ft-role<i dlfFcis iila^htly for the aiL^lyjt'. nf 
WorlpinllLalli'^l. Th*- eKLrni.^l fl-rrnle, ntlln_-t lli*h UflafcR iiiii-iftiu'd nihthbile tli-^ jin^li- 



attpf^e prcijF=ctiors ift Uten I n fl-ppijur I ii lh<>EuhJLM:l i^wtliriii, :iiitL LkioroJcrc U u also 
■ttJtt)gfLi>1 niternall>_ JqIetHhiI V t^\^ trd aasiRinl wiLtiin tfni [nfojffl' Lloiu *ird In fact 
appear ^ ot>JKtR- 

Tll?T* LR DDoquir^lion Lh:it :iryi^4i: why tilk aboHl t^^^ gLil^JMt/ohipct d lb tint tLm 
when there Is ;ton{Hto-one mippin^ ^^'ith Iht^ ti:tefna|/lriLernat diEl][n.Lioa rct[i[]TTj 
by P TllfflTy? That isn tln^ cLOtl<J]l« fif gubiect ajid obji?i.t accm to h*^ rfflmidanE. As 
WiLliamB[VViL^4l bi5 |]«utrf ouL, the ustonial i^-HJe di>i^ not always map to tha 
tliwJfitr pmiLiDQ En Ihf Vfjh phraj;? {WarlpirL'fi ttiliJL^ct pceitlon)- In Eapjishi far 
OBiiLple^ tht *til>j^t nouD phrase is analyHn^l flh tht spt'clfier of the prcjcrlioil d 
INFLh While in VVarLpin tln^-^iibjecl do^E, indi!*_^l.^|j|tKtr Afl Hn?_Rp«itier of the ^erb 
phiiue, WHijie^ LhedLatinctioa bclvrtMl gffunmaLiLftl function ai]d ^^■^tl:r]laL/JLU^^Ml 
j-nnle m ordur ity jnftlPt^Ln cra9&-1jii^iii£tir gfner^lLEJ^Iion. 

As irt ftXimpli? l.-flLnicture. toJbSldcr like verh stem, pui+fa. Ufi L-*tnLCtiiT^ in 
ibowfl In figihre 1^-1 fi. FoLlowiJi^ the nil? ahovOr we nule thai, it^ agent fl-To|e w]U 
appear as the kqiI^JiWIt ajid that ihe oihpi^r* ft^lll -xppfiar as objectB. As flhuwn In 
tftble?-^^ the verb Btom itMrlE WSLgTifr hoth the afjent and Lhcinir d-rnl^, Ki?rre Lhelr 
appearance in tliC verbal projection. The JtvcL at whtch tlli?y appear h dicUkUsI by 
iLctr gr^ni ^atical luncLjOh. 



© 




"^ 




.^CEM 


X 


/< 


\ 




/ 




^ 




^4JiiJa 




TUEME 



Rgnre a. la: Thw L-Elruttu/e for pziti^-i. 

MJwIns ^roni the L-Btmrtuw fiDr punlfl fa th? path ^-role. This arenment b 
pagcuiHJ by the flattve case- marker within ita 0*^n projpctlci]^, as shown in 1iEQre2,]6, 
Note that wh51e thfl app^fyyi^^- In aynLaj of thi^ roJ*i is lidJ-'cisetT by tJhe predicatoPf 
thf ja.?j?hjFtJ?^enl dT its role is pnirformed by a difFeiont elenn^nt. The two ■^triqctura 
arc incler^d hnled together in syntiM, a* On? WOtild expect: this iadiscUssOd hi?[i>w JA 
the£*cticn oti D-*UtlCturf- 

ThoreJB a well- formed ilCTH CCitlSt rai nt iu the mapping fiomfl-^irfs ICi L'^tfuctnre, 
itameLy, the fl-CiileriDa (fahen trotri ivltWRfi]^ r:/-^ [ChoSlJ): 

S^Critcrim; Evcfj thhlii [l.rn, ff-ponlLloo WBK] muat rttcivd Me UlcE 
oidy one 5-ro|ti [p_ ai5] 

60 




Figure S.lfl: Ths L-ttnicture iw" Oat. 

Thifi prirftple Ruirinicof that n^vf^ry ff-posLlaon, as deL^riniiied by tliG l^;>:kfil 
llifoTFnaLlDR asBociaU^L wUli the predLcit^r, ^Lll heUlkd B-^khafl-rok. It furtliPiTnore 
EuaTititWflth*teajiLBiJchpt^hjtT]bi^lllT^rrt befiUedby niut^itha-nonefl-ri^^^ OLwrvi* 
Ltat this priptjple lia& b™il Oljeyed in the L-iLriH^liir^ fur pun^a and DAi"p*lw>e- 

D-StructUR ia UBed Lorcl^nscnt tlie predicat^-ai^U tn^nl relatiDciE of a Mnttn^^e- D- 
Btructnrcfl axe fornsiid by COJTlbmin^ llie conAUtUtnl L-st niclurtfi of tKn^ predic&tors 
aad irE^iJinciLtfl #- aaaignmenl and ijLClir«^t #- as b i Rumen t licens*^ (^ombinalioii. Not*? 
that tbf^ licEntiiiR TcJatiocLfl MT> theniBclveB lit^cif-cd by Lhp BernanUt tontCfit-of tbe 
prEdiciiCLw. Ttia-t i£, the Bynficlk f^Ullrtn for ME]|;iiinE ^ gu'fis ff-role miy nul b< 
pffjf^L in Bynta.^ urL&s thfl p r«5 uiitor selccth tlitt fl-role. 

Forid OLampk D-stmclura^ LCTia id^r tilt gample wncirilceT (2)^ once again. FLraL 
wcexaminelhecoreof ULFitiHjnterinre, theL-ElrutC^JJ*(or the verbBlcm.pif^^ 
ibove in ligure 5-^^- The verb tuilwtb three fi-rjles, Iwo L>f which are Lcen^cd Tn 
ihe L-atrnctur^ by [tn^^nft cJ t-lie C-aBEC^ninB fkjncticrafi of Ibe IbtnJ ^ui*l firBL-Jevcl 
projeclion- Tlie llisrd is lirenEen] bj- Ehe daLive caac-markor In its L-slrucUiv, M 
Bbown in J3?urfl 2.16, ahcpine- The dative L-etmtture ia licen5r]d, la lUTPf by Ihe 
indifcC^ fl^ aligning Junction dI tliC verbal hcaji_ 

FigUfe 2.17 HJeplctff the D-ttrucLurc for ihe -sample seilttlitP- Wp see that tin? 
ngn^nt, tij^ajt^N-rVu, iB attai-hrtl as nlbhnff of the fint Jct-fll projeaion; Ib^t tlift theme, 
iarJi, [3 iittifhed as sibling oi KW' -r-i^rljaL head, and Hiftt Hie path, iurtlcJ-iHT la 
Srtt^rhed aa Bibling ol tike dative caf£- m FiJ-Tcer, wlikli itself ia attacbtd ft5 a sibllnR 
of ihti vcrl>il head Kvecr L-alrUtlUlC has hpen properly LuTCr-^ed b«:aiij,e GOfh hftfl 
been incorporated iaLu Ibe fltructiire. 

2.^+4 S -structure 

S-Elrntlijrt repreaentB a different fry ntartlc view than D-*tructiire,eMentially that of 

caae-martina. Ca*e-cnirtlnRaBKKiit^» prn^nliriLlea with their arftumentB aL Ibe kvel 
fif S-5lri3cture. Tliife lifLsnglnR relation ia r^r^l^^l ^^ ^^^^ lelaUoii of S-aaaiRniilen I . 



ror CMC murLLaE p-NcntTbCrA t|i*l Jn involye 4d Oti'^ h>i^^ mark- t Aliftr^i riio u atolLmrnl tii 

icnilflt (nr EHWHiaatJVU ^hJ J. CII»*Um ciM-marlii tt^ V-li^CC Ul« veih ah£I^IL« Pii;m i-^i*^^ Lu Ll 



ir 



fii 




r3gUre2.IT: TheD-Birnclureror wnLj?nral2J. 

This Eetlioi] b^fjnpi ^'iLh jidLacii^i^LDn oECAEeLhe^. AR^rllus IsftsecUoD dii tlifl 
plFLCGU^ent of cu^inarkkn^ relationf ip lIiq EtTucturc. Lazily, the s^cliort diBcuaHca 
bow u> rtprnsejit tJi^ AimJinri' iu ti-atriitiHure- 

At S-&^njGtiire, ar^menlE ^tg aasaciaEed ^'IlIi rheir preduiAtes tie c&jc As with 
seiuatltkc»oEcctiE^il 0[^-r4>]«, [Hfixlicalr^T^ qlE^t^nclin^ tEidr r^^n ^u/^^Cr^cprr^fTCPrt^i.fL^ 
LhE casGs thzil Lh^y liceDE^J"^ TEie^et of casea Tor wbich ^ pred]cator &ubcatef^TJ7Cff 

EvaLoctiL cafCE^ cr;^t[V€, ibsolutEve^ ai]d dabivE. In WarlpjrL tliEre ^re li^'e grciiips 
□I^Efbs Chat ba-ve diQEienL ca^e liirjyb^ htiQwi] ii] lable 2.4. TIiEiie live c]^5ei>s [Lre 
H^rlv^ttve fTVrf\ the ll^t ofpoAsiliJ? l^-^rkdn Tnund in W^pLri. 

The ^-structarE a^scciaticm of ^r^mentB w]th prEdicalon is sBectEd ih a tWo- 
part relfttkm, AreumoiH* — which. Tot this tlitfllfr. are only rn^ima— arf aAsotuj-ttJ 
wjtti a EzaEe-markEi by thE relzlicm nf c a£p- mark] n^ Ca£E pljraA€£s lq toin, ais 
MWKifttwl with argtarciunt [nwlt-lons Iti th*^ prf^ic^tor^s prujwtifhn hy th^ Tn^lation of 
casE-a^signmciii. 3<3\b o[ lhcE'> rEEctUonfc^ obt^n Under tha TtJAtiott of l^VErUm^t, 
aa with 0-aaEL^iuent bi U-EtruEtarE. 



sub|ect And obj^ct^ icapccii h'd^. I'Lc pAT±cr nfiaakf be ^Ll? tG LacarpiDrTLlc ubFlncL cbdc bf ^tlji^ 
cojc marLiJi^ cvp^lnliLlci Ld ]fucaJ iE^mn DlLrr tL^o jun4 cutt^iBLrk^r^ 



B2 





case 


arra^ 




ejiiLmplfl 


L. 


Abs 






Jtf 'ta fiO^ 


Z. 


AB3 


&AT 




^(lU-a ^to [ijv&' 


3. 


EKG 


DAT 




Tfffjrrc ^t* -^eek' 


Ar 


tftc^ 


A [3^ 




fljfl ^lo ett' 


5. 


EltG 


ASS 


DAT 


j3Ln^4i ^Lo taLe' 



Table P_4: The Rve^eibaf fajie arra^'Sr 

In Wurlplrl HiC^ir Sirt Lhn^ Sjnla^Un- LBiG- mirtei^ [tllG. AH5^ and T}AT\ conc- 
^pondini^ tc each of the i^y atactic ea££±' ^r^galive^ ab^oLuti^f, and diLive p'atu rally 
Enonghr each ajntai^tl<^ CMc-niirktr marks its atgurcnint nouti [Hiirai^c^ for Ite ow[i 
C^tfQ, The casE]- m a.rlerj and tbeir phonetic realizatEams are presented in table ^.i^. 
The list of WH'arLpiri (^afle-nisign^ra Is given tn tthle 2rC, (^Ot^^ that 'T^ 9ta»dit iQr 
'ti!n*t H_-knlflnt\) 



ERG H?rgariV0 -ngtu, -n^ib, -r/u^ -rii 

AB& abEolutive ^fi 

DAT dH^tlve -lb, -*w 



Tlblt 2.5; TIlc Warlpiri cj^fl-marknirs. 



ui]gner 



V ergatiie 

T alfcsolijr.LVj.- 

DAT claLLve 



I^M^ 2.6: The WarLpin cB^-a±.fciJgnertr 



PlgUTe7.]S i^lLc^ii an OKajhple njlthe case- [narking [)f an trgatjv^ marMd noLin 
phiase- ffiu is the caae-rnwkfT and rfcyajnicj is h^ing marked for case. Note that rJu 
f3i>4?«. md^H^, gOV4_'t-iL its arEument. n^ju^u. 

P!a[-ii]g the Arguments 

Case tliHwry dittitt* how caaei are IjceDi&d in S-Btructure and how argnmentj^ are 
manifei^L {t e-^ LhrougJl tin- r^liktlons of caie-ccLirtjn^ aud cafte-aii5l gn mpnl ) ^ bnl iL 

fi3 




dodS ilCHt iddlc^tJi wbttt tfity ftre lo he I^J&j^j^- SpwJflcilly^ lL doea iidL indicaLe tbe 
^ftincn^tjcE] fnDcticiaE correEpandLii^ to lIid raseE Ik^L bav^ been LirenECfl. Tbrri? i^ 
^ ain^ple R«L ot ru]£E chat d^teTTiiln<=2 th« mapping betwi»n coEe and j^r^mmaticaL 

flint tli>n[Haie3f. 

1. Identify the subject functDcn wiLh chs EAG ar^ujQEnt^ if there ie duEs ott^rwise 
wiL]> the aP"^ argiament.- 

?. Id€iLt]fy the object FrnicLiaD with ilie DAT LTf^uniefit^ if tliEre Ls Dne^oLheni^ 
ivlth tht AB5 argument (If tlil^ Is not Atready IdentKirt] jm th^ eulsiect)- 

Fi^re ^.IQ jJhCiWB Lhe S-sirurLure for (2]x'^ Obflervi? that each of the ddud 
phrafies hu bcpii n!nrk*>d Tof iLj catfl by the appinpnfttt ffcsi^-]rtarkor> hy vlrtui? of 
ibeJr^veiDed tt^lus. Obia«rvi>fur^hsi that thccaiepbri^s h&ve bKaiAsi^iu^ Ibeii 
l:^^w by tliq 3ppn>pri*Li' t --wLj-^j-jgcK-rS . For ibt^ iTfiatiVr Micl ihsokiivt? ^tguiht'nti>T 
Ibe Eis&ignincnt ^^5 pcrrCHrcriL^d b^ Un2 lirs l-levcL pn^jt.^ UoH o( Iht verbal htad itld tJin' 
ten&e elemenl, respEcti^ely; for the derive ar^mer^t ^ tbc EUsignniEnL w^^ pEiformEc] 
by thfl (lAtivfl i^ase'ftJi&lgTier. 

AiLxiliary Syntax 

The ^ynt:^ i^iE tb^ itL^xlllfLTy ^ sot^\^ysi\\kl afi /id^EZ^duf to iii bLt-und^r&looLl Qalurv. ltd 
m^ tniictJoii j& lo combLup w]\.i\. th^ n^t uf tbi^ ^tlli:]kC:G itl tvA^ W^y^; Lt COmhJMH 
wkth the tEHEe of Ehe verb to add afpect imfarmaLkQii^ and, it combiDes wlb eKe 
iTEumoiLts for tltP piciptw? of pejfrln and fitinibcr i^ji^m-iDt. Tlni HttilCtUfj.^ of the 
^lUJliary J^lJlL^tfE Lhp c Oil q bj Fiat LOU (whicb t'\ ^HscuH&Ed in lbs jccLioi] on ^Eiriantic 
iqterpretztion, below}. The base 3b conBjdercd to bE LbE tiEad of thE auxiliary. TIig 
noniLJiri] 4Lgn:i>iiieiil rliUti^ 4ire tnh^ to be itA aLpJfittf, ^id thf verbftl projection Jeh 
labi^ti to b^ ]U fitib}£ct. A& ^n CKiLmplc. this compltto S-atructurc for thi^ s^^ipla 
SEDtei^ce u depicted in Jx^nre I^.^P. 

2-2*5 The Mapping Betv^een S-etrpcture and D-Btructure 

The InVclfr of D-Blrartun? and S-aLnjcluiie r^pu'wciLq tl]H_" sycilaX uE ^ SCJitOlicH: frnffl 
l^c dJflerent Fupecls D-strnrluTJ' is roncernod with as^i^xoeril of ^-roleh^ S- 



^^Tktt ^mMiHwiy pmjtiitwpn r^ tinL riMrWN httt^ aa il wiU 14 Ji^iJE«^ bidov 



M 




iijjd;vJ« rFu 



jruifJo-nif 






PfiUT* 2,15^ M-M% <Jtbe S-ilrticlurt for utfittutn^ (2). 

strunur^ r^lafCS^Dta mcc-niarkhiiK t^IelLjcile. Bolb af Lhp^e levula, Tiow^ver^ are cod- 
cerned wLlh tlie relitlon bet^^n II pr^tcatn^T Mul \U ^rfUrr]<-Ti Ls :i^ rrUUliEf^tod tn 
syntax- Qa po^jls tliit lliC two [evds iJ^ J nd+xd relatedn aiitl Ibat one structure i^aD 
bE] traji&fornied into Llie mther by moveihi^ffiL. TliAt J?, rii^i^ Lliit CKiJt in i^n« [i\Ra 
QlD-atTucLkjT^ in OThI^F to r^^ve th^rfl rolnS ni^rV nn^Vein the slnittan: art Ordpr to 
rtinLv^ thn.' corrcftpQadin^ ca^e.^^ i'htfl LcBiisiDg r&qiLLremeiit for axgumep t£ is ^veq 
by the Caae FULerfCKoaiJ: 

Cmc FiUfr; •NP^^ if Nl^ h£ji phoiwiHc wntMt uhI Ism no Ciflt [p, 49] 

Movtuntht i? ^iUqwciI ljy tbe very sfciriple rule of Mqv^ce, ^taLed in (1?) (Laken 

Eroid |vnwse|. c/, |rho3i|). 

[121 Wove any catiij^ory a anyivhEre* 

Thift ru[« niUfit bt r«4lrict?dr (iow€VBrf so il won't ma^tiVely oVeff^Wit* un- 
STctinma-Ljcal Beotonccs- One atjcijig ronytr*int 1* tJie StrLii:tnr^Fr«crvina Hypn^th- 

jEr[w76]> 



Tlitt? arc ulLcr Tco^^nn Fcr inuT^Di^iLL, Hjrii i:v M^iitf>vi'iTKqi Tt^nir ptitnuiihrDa lit ontnittr Uir 
^rvi«W ijf till" pacHT, SG ibcjv waa'L be covcrnl hrrt 

^'^K^' IX itkc tjfti^LUOMi] TiOUtlun fur a auua phrur ll oriTmpDadi ta CIk muirru] projfKlLofl of 



59 




Flpirc ?-Jn: Thg S-nriftqrfl for lusntencp (?), 



6fl 



ur iiLbcrtE a D ide C iiiti>sonLt |nJftitifcPii whxre C caii be i-cliawlse g\ JiEr- 
aled by the ptamicir, [iJ. 3J 

For tlbP iicnpLe pheooiiieiLa h^H-ndJ^I hy Mha p^raer^ Lhu reiJLy hwls dc^iitji ti> Lfai2 
movPTTiwit i>F casi- ptirrfcHffi- Fdllowinf; the ron^lrWJit, tli-i j^irbGr shoald Drily allnw 
laioveinent of case phrases Into H-elJ-cJi ture positionE that aiplirenMHi fit D-alruHztiire. 
Rcmfhlj afreaiinRi ari^uTnents nw-y nrly rrWiVc iiiLo argu]neiit pcfiilion^ 

Anothi^r limit lo mo^Einenl b Ifie i*m]pctlO[l Pt-LilcLplcn frbkh cDi]£traiii-^ ihe 
pcefllbLjj Tnappii]RB between the ftrgtinn^iLt pctiUons of S-structure, [^atrucCUKr and 
Lh' (borrowod from IvHWtibJ; cf., [ChoS]]l. 

Projc/^lion Principle; Thed-CrU^riar holcl^ ll D-&lrDcturCi S-atrftciurep 

&nd LF. |p. 252) 

Tbia principle ensurfS COil^i^lOiLiiy bethveeii iliese Ihree levels of KpresttltatJon 
[for the p^tn^cr, JjeI tte levels of n-eLiiJCtLire intl 3-*1rurliir4?}. It estahlkhea n 
CDonectioEi for e^fi <^f tJiu pnidic^lor'fi argumfnt pngUlons at CMli oT the Icvols- 
Civen a f*se phrEwe Ihat has hwn ijuignf^ ft fl-tolt il3 i cerUain pcfiition in D- 
StructHrGsil y^ould be Lncop&Btent for AJiutLerciMphraBeloniDve into titatpfflltlon, 
tn & BenM usurping ttit rt-r^e iBsi^ed there 

In WarlpirL, lho^^e\er, Mi>ve-<^ a^ rarery used. It Beems that the parfurieLeriKatlot] 
of the LiJiguaje 15 ^onstrikinCTl so thai Move-i» need J"pply yilly in ^ ftw, select 
lllfltinccs. As for the paTfter, tlifl siuiplo &H?nteiicefi in its domain rlo not call for 

MovHZ-Q at all because argLimcJilfl n^»ed not move to recieive naae- the pcisUiorLS wlirn: 
they are asii^U'd thpir tf-roles are the flami^ pnKltton* inhere IbtJ ire assigm^d the 
«ill«pO]kding ca£€. The cap he seen fOJ- tljt [£i:ippii]£ from ^-joles to caaes^ shown 
In tahle2 7-^* 



?-ttdJi ciK- 



AGPHT EftO 
TLEME AB5 
fATM UW 



Table 2.7: The fl-ro|e/rase mfl.]j]jiiia fur WdrlpIrL 

TtiDs ni^vi?niriLt ti«H?d not enter into the theory fm whkh the part^r k h^u^c^l 
Hfifausc of this, ihe S-struclnres Mkil D-:tCnictUr4^ lo^k thn: samt> MLLcpt for the 

nyntaS of the auxiLiiry, whiuh ui rcpnaeiiled only ia S-slmcture. 



^ I' afcitith^rel J-. tbcTE u na fTflaa^toty thflorj for thlfi hupjURg. The JwcfiL^tiv^ lliftijty of Hiiii 

nuppjTIE ^nmpL^ due LEPCDiprLfi] utu^ElH 



3,3 The Lexicon 

matLOD dirlatiiiE clLctnnitlJrtfstriltotlR Ifl bolli precai ent e and Ejniadlif fiTfiLCHir^^ A 
comfHlj^te cntr^ H^ontaim ipjormatioii Ihal it boLli hp^jcillc XO th^ Icxj^aJ ileniT aj well 
a£ ap pLiriblo Uh t-bclfedrif clawf-fl flf whi^b it i^ a inembEr. T]n:lcJ<3cOCLCOftt;4ln»* twu 
iTriLrtijh^ (of rf-prfifionlLn^ each kind at Le^Jt^ []kfomi^l]oiL- 'I'l^f^ tint je a mappcDg 
tr<m] itEinfci toH^nUie^ corLtalnni6K|ier_ific miormatiDOj and Lli€ Bi^^ftild Is ftS^^Cotlij)^ 
tiNll apEiUng lo cl&agfti 4irieKLCftl iteniB. 

Lexical ^Tllrws CdntBin calsRoryt preced'Ticc, KyutactK, *Ild aenianLk ]Tifarina- 
tion- CalERorial infHjtEnfttlon rfl<>r3 tn ih^ Ic^iical itJ^m's piirL of y[M^Cch. JcicLuii^d Ln 
\h<z sfi>l>i^ ol tfi^ p^PW are nouns, nun] be r-markEr*, ^jnloCtlf nate- maxkeiET verbs ^ 
teosE ElerasnlSj and iUJ^ilJa^ j? i>ni popeaLa (haaet^ nuhja i hL-qJ i>bJKt clltK^n fl-ud 
dc^tlve r^sLiaiioa niarLersJ. 

Tiie precedence and EjTitairtk ci^m p rimint^ oMf^^Ica] eatrLEfi dicUle tliEir piTtlfl- 
pAtlon In th^ i^orreiipondini^structuret. Tbwr fttrUtturefl al&o reler to tbni catflj^orial 
mformalion ol llieJ^[Lti>. For t:^ainptpT rafae- markEra may only combinH^ w|L|i uoihups; 
that ia^ the catsfioo' oT the cocnbinGC [« rcli^vant to ]*S, In SSr ll»e caleacHri" of the 
UgiB d Ctf^ntlJ res tli? number of levek o( p jojk tivn . 

The la^t romponfinin i^^manLUi LnformaCioD^ maaLrestK jt^£][ E^Oth Itl s^nl^T ^nd 
in semantic interprElalEDn. S^niailtjcS fs totint^led lo syiilai ibiDUfib Llia ntippltig 
offl-rolcs Losyntattlf tftt^t^ri4^T a^ w^lla^ caBcs (dcsciEbcJ itbow}- Thi; peison fl-T^d 
imnihyr of nounE, pronoaiis, :iiiil asudltiry ^it^menl clitica B.ro prime esampkH of 
Lilltr[fcreti\'C IflEurmatjnD Btored ja lEle lexi-:on. 



2*3.1 The Lexical Eiilry 

LewrJLJ Piitnes rontaia inrorcnallon for tatli uf lhii four components listed above. A 
Eun]n]4try of t^^^ I ncllvidu^lly Stored InEormatJon ]e j^>en La tabic 2nSn Not^' ih^t caLe- 
£OriAl infoiniatlon ie always coaLajncd In Lh€ iadlvldUAJ entry be^^vs^ it tB partknlax 
to tbe Ic^cal itGlh. 

Ths prEcedEncE infonn^tlun Sl&red Oil a per iLem basis k rathsr i:naaHr Verb* 
und tunfie elf Kienta contain tbmrcoDJa^tiOn tlMftfotlC tolivej, wtiichis med daring 
lailcM.tEi'l vi^rth Analysts. A^ixiLlary bsaea conLain thoir Hucubi^t' oE^yllahl^^ lia^d loT 
(i^OckkiLg tbe well-Jbrmed n€ss oE i^'ordB. 

-SynLacLhi information conceniB btitb S-fitfiiCtUW ftnd T)-sLructLire_ The case 
ipajhed by a case-markL^rdcpCJkds Oil tJlt partlf'islar raas-marter aad iO It li Stflrcd 
][l llie entry- The case and *-role assi^Dcd hy tho dativ* cftfle-inar^er, unlike other 
pcodJi^AttHnST is also stoiwl In the entry- 
Tin? aertiaiatli: component Je also at rai^ltfor ward. The pErBon ^nd Jinmher Iftfor- 
matJon for pronounB 5tud f^r^jncjht tlJtk^ depcndB on Lhs parUrnlaf kxical Itmir 
aa it tB stored indi vid OnTlly. Thir tAm^ 1^ tTU^ lor the tense ]nrorii¥alJon oT tntti* 
fllBfOentSi Wid th^ wpect of anxiUary baaea. 

OS 



piK^dcnce cc^tijiLgSrtJoih <h^^ (v'^t-E>$ AncL tense fknientt) 

ayalactSc caio-mij-lcitig (j^ajHSn^arkcn) 

caAG-a^&L^iuneii t [dative cafic-inarker} 
ff^^sLgntn^CLt (dative ca:te-mfu-ketj 

isnflntk person fproin^nns an<| ^reement fliUta) 

nijEtibcr 1 [HRinOTJiii^j n i j EnSrer- rnHLrkenh. agrtcititnl clitkfi) 

tens? (teriHC demeni.'i^ 

B^p^ct And teiLse r^LrLclionfi (auxiliary b&sea] 

Tablets: TIle EE^m^btri bTui Sudividtl^ lexlc^ 4tltty. 

a. a. 2 Lexical Rules 

Ipuuc^ rules encode dEEcrLptive inronnation that appE]E9 la cla^Ees cf lexical ] terns. 
Gbch ruk is tn i glmplu '^if-then'' fojm. Ih^. cunditlcmAl pan ifst* fnr niemb^rahlp 
In a f^ven clasfi, and eIie a^itioia porL indirates the i nformaLhan to be added io the 
i:^m^Utj^-^ l^K^ tQUj if tLfl ni<:tntH;r?h]y J? ^^tisflirf, TJi^cl???^ of inf^rtTAtlon 
represEnted with lexical n]|eE ars lifted i^ table 2.9- 



prwtdtncp adjw-Hruty rcqaiifincatft 

direct^ ^reument identtficalion 

gynUctic cast^as&l^nj^Dt (prd^nlicAtorft) 
^-asH]gnm«Tit [prediratora] 
^'liokid^ {pi^diraton) 
projftcckpn (predm^ttoff) 

Ta.Me2,9: Tli^ r^ngH^oMi^i^lcftl tuIpj- 

TJlC adjltcjl^J rH?qulreEnenl&rcii tin? irtrmuK ciitEgcrits are liiltd bdou- For mo^i 
calcgoriEa tbE re^iuirecnefiLB foljow stru^ti tibrwii rd Lv fTom tbe data. Tbe Auxiliary 
jeIi^ on the mow luvolv4>] notloja of i bdtrir totnplatt. u JwcrltMjJ above In cbe 
BECtion on P^. 

' NtUnb^r-markErb must be enclitjc \n uoude. 
■ Cai^L^ marbers zaust beencQtlE: La qduil^. 
* TsDEe EiEinents must h^ enuLlIk trj v-i^fbi. 



* AuxtlL[kr^ wnrdfl may bn rn*"-lUlc to any word 

* AyslitarycompcmGialflrtLilHthH^onclltLc to Kn:h other uj^LH>nllliatj> the t*nip[atfi 
^Vpn ifl Hgnrv ?-fl, 

Tbfl hmIi" LnstAiice oJ dircclwl cU-ftUtll^Nt ideiit]1icatioQ hanrilGCI hy (he pfirfter 
CKCuiS whth ciBC-marktrt:. CasiV rfiijlcei^ take botb Lh<i tiOUtlS to which the> aie 
enclitic iod the prfUfldiiiR fioude witTdll their i^hiMe ^ arjlumtuti, Thl& l» stated 
In tlni rules below. 

* Cpsp-markers K^kn tlie Indians to which tliey arO HitlcUtlc :i5 JUf^umenlt. 

* Cftfle'n^&rl:Gr& t^ke pt4>£L^lng IiOHHf^ ^ ar^iDei]tfi. 

Tho ruW iVn r3fle-H£5]fliiniF][t ajul fl aesignmeiit wert given in the ttjrm of tables 
^ihove. They arc CJit^ocled Ifceie iq the Ih^xjlOeI ^ rules The other conil^Oiieilt riE 
t lie syntactic ii]tormatLi>n rfKiaU^ lhe LdvgIb of prtJKtLon for each cat^orj, li&lcil 
below. Vn^rbfr l^TOj&it two levehi Ln Otdtr ta iireaL€ slots for sicbjo^ti* iUlri objects. 
t^&se-mark^FB project Ojke Ifvjil ss their only irgunients are she nounb thit they 
mark. As mentioned a-bohrcs aRj>JllAn« pTOJect two levels^ dUe lui the a£jeem«nt 
diticst and one lur the*rftunie&t verb phff^C. 

* Vfliba project tiii-o levels. 

* Casc^jiiark^ra project one li3VtL 

* Antiliary basce prfljwt two LevelE. 

3.4 Semantic Ititerpretatiou 

Thd i^tnanttr mterpTetation perfr^nne^L Ny the paraer opemtei i>n botti 3- ar<l li- 
st ructu re, d eptn^ i fi g f>]h the type olinLerpreL-aHon inVLJ^'ftlr Kor example, arj^iimQiL 
idcntlliHr^tloii ih read o(ForD-*tt-ilCtiirei itslng tbe fl-aMiEEimciit relation. Agreement 
witli the auiitiirj EH rhn^ted at S-&trucLarc, ^ the ai]xlhAT> is not repnsentjjd it 
D-Str^t^tlJ^e_ ThcB section cliStUftSf-fl the Bemaiil]c intcrpfetatLOtl within the docnain 
of ibepiracr; iraument idercilic ation , ] n t*r|]ftt At Jon of mitl elements (nnU mxiLlEtry 
corp ponentB and duII ^fnvphora), and othEir BemAiitic "veil -lof ped nesB conditioiia. 

Z+4.1 Argumedi TdeiititicnMi^n 

Arrument idcdtifj^JAtloiL \^ the asBocialicsn o[ surlaM string componentB ^Ltb tiia 
iHini^ltEi: runct]onB that tbey Hll. ,\^ thia interpretatioiL cofiteJ-flS tlLe Ren]:iiit]c as- 
p^t o[ tllf^ ftj-ntaftir Fitructuraj D-h tCUCt Uire Is need here- ArgninenLj W^ IrfentihfM] 
bv virtue of ha^anfl bi?^iL AiSlj^CiJ*d a d-role IkeoEed by (h^ prertKator. ^iirne tbia 
LilforntfttlorL IS repTPBentnd explicitly, thft intf^ipretatLon oF the i^'CLtattli- atruetuje 
is Etnugbtfor^'&rd- FoT exaiiiplen arguinulil i ft terp relit Lou for the Bampli^ fleoteiice 
^ve:^ UK the re&ults m table S.lDr 



0-ro|E WOFlI 

AGENT nya;u:/u f'l^ 
TJIF.U£ kfirii 4'l3CiciinEraFig^) 
fATff tirrrfurtlLitiJ^J 

Table idil: The tf-nsle/wnrd mapjjlils for (2>r 

2.4.7 Null Auxiliary Clcniponent^ 

A^ nii?llLlanE<| In the hTitroduction, diuy parL of Lbe au]d]iary t^'oid iti^y bo pliono- 
Ic^uUy nail. This doot doI cq^il thi^l ^bi> ^orD^lhOcldtne lllf4'^mAti4^ii \3 mi^smg, 
ho^'^ver L^ll bu^^iltary compoDeDL baE a delault vaJue LbiLt l« app1l4>d L]h ttiO %b- 
SEEHie or overt morplieinea Lo Lha ctmtr^rjr Tke JkULI ^UXtll^ry baie, foUowiiiR the 
tEkble pre5«ntEd jfi tbe inrToduclory cbapt«r^ indcrat^ ]^?tfcctlve a^ptct. NqLI ftHK^ 
Diary a£r«>incnt cHlita have dtfimLt values ot thtrd-pej^oD , fringukr. Il tbould be 
rotFd tbat their mterprelatLoia dcp^iidh on tKe VrJ-fc's fttlhtatH»gorlr*lH^ri frFimc K, 
lor H^K^mpIc, ttic Vi^rb dCK* llflt _r u trfategi^riM for «^ object, tben Lhere is no dtliidt 
Enterpietatiofi for object agrTemelit . 

2^1,^ NuH Anajiliora 

In Wartpari nonej^T thr itftutn^CitS of a verb need h? expressed by caE€ phrases; BO 
ATgum^nL need be asBochaled wilb ai3 ar^mcnt poaiLLi^ Itl S-atri|4^tLqre. Wb^n a 
ca^e plira^e is abhtnL from the riUfliro irjtrinj, the corT^j^pnnd ing repBLratiDD clLlic 
In the nu x1 1 i :^T>' tA.L^^ on more of a pToiLomiFia] character^ Sucb clltlc^ V4>iild b^ 
Lianslated ae T or^her,' for eiampk. (I:!| RiVH^t 41J1 ^^aitlple B^ntenne In wtich none 
of the irgutni^nts are overtly eKpresw^ ^ a jyue pbraBe In &oine contG]^!^^ lliir^ 
would he tha prelerre^l mode of expreEEiou , and LnEcrtfciLg ijVh^I ptOUOUQ^ would ^ve 
U3 cmpb^tic rn^j^lu^ 

'[ ma> take him/her/it from bim/btr/3t.^ 

7w4'4 S^mantit: Well-rorni«dne£s 

TbcTtt arc Ibrciz ccmdiLiona On *rl]]anlic wril-foTmednesa toncemiliff tb* lit«L«lrtg pf 
hnd agreemeJ3t ^'Ith mnipc^nentH of the aiLxLliar^L Note tb4it b4?cau$o thn; condiHonE 
involve the aQiiliiry, lhi» itltn^pn^tadyn as ]jerfurmfd on S-stmctnre. The first 
condition^ given m [I^L), if^ latber Eliaigli t-forWftfd r 

(14) The aiijd]inr> bi^e rtHJit hf foil] [Hail l>le with thr^ tense of ih^ iLifleclcd vorh, 

A&nni-^ampl4>ofagnE>imalltal uftoof LlifTtuxiliary, consider Ihe^ctip^esr'ftience 
Otice agajij. Table ?. II itbiTW^ the le^tse correspondence b^Lv^n^fn tlie auxiliary and 

SI 



liiixildary^ tllu tfilli^n<;?* is roDfiiilr^i'il iVh-I I ■ rrnrfnod from LhL (^uLnl irf vir'w_ 



fx>m pa table 
base l4m&4^ 






t^ DDn-patt 



mt 



PGR-pafft 



Table 2.ll[ The teiVSf fj^rrespoudence for f St- 

The ei^i>nd cw^dition concerns tliC lU^^nstiij of the ai!jcemi_-ilt cIlTK-^- Aft wiLb 
MftUITl^ntflT Rbe semajLlif ^Jgument aelpcl lor of ibc prfldlratm" dinLite frhich clatlti 
ftre UceiiB«l. TbLs It: sl&led m (Ifil. JMoLe Ibil agreement cUticB are b^^\ formulated 
In ttfjug of gmjiiiiialical ktiCtion— not case LendLne [doty^ sijppoH lo the coscepl 

of firaraniFitbCaL [unn^LloTi fin addiUon to ff^jitr^J Tacta as nuralLomn! abovn*}- In Fmi- 
rOtrlini^H* with the Exteclrf^^ Pri>j*Ktii>n Printiple (llic reHjLLirement cliat predicate™ 
have subjects}, ULfl^ mn^t alwz^'E be it Stibje^t, Kf this igfecment cLUic Js Alw;^ys 
licGusHid. rhil^T tbi£ condiLLon iWilly &i?rVH?s ^-^ a rondUaoii en tlat i^ppearan^^e of tht 
ohject clitjc and Lhc dallve repatration marker. 

jlS) Ncmin^J ^^sreelnl?nt rlLtlca mnat bi^ Jittflftftri by ihp maisi predicator. 

TK* Mt cmdiiion, praictil^i in {10)i also Jollowi from ubseived rfj^ta. For 
examjile, consider thf apeecnent correft|iD[ldorLf.J^ M the &ample fienlence, HhoWJi In 
Lttble 2-L2 Thia &eiitence it ?i[so i^'elf-frnrmed with reaped to Ud£ COJlrfltioFir 'I'he 
aabject clitin:, nidr H first-peraon Ei]iEtil*w. v^b'Kh agrf^ with the subject pronouftr 
fijflji^Jti The object clitic ia illllL ^^^ thew-tore defaaLlt to Ihlrd^p^wnn aLp^:u 
(fta dsStrlbwL &]?Ovel, which aprect WJth tht object eoun, ^rJu. BeculaC iwnJcJ » 
HTi marled far nuinbtirj it i^Kftes wjth eithei sanj^ar or pLurtl^ th^ rorreRpondLiiR 
fUtic dJ^^Utl hlgua^tea betw«>D thr 1^*0. 

(Ifl) The nominal tpf^mprt clitLca oT tb^ ?mxllLary must af^r^e with 
the ugumeiLtE cfthe main Vtrb m pcfsoq and enmber. 



GP 



clitic peiicni/tl timber 



Br^amunt p^^^H0n / n U mber 



fllibhect 
objwt 



ma 



Hiu t -p^_'rsoit alrgular n^jnJu flffit-perfion Ein^lar 



jjr thinl-pej3on ftinfbUr /^ntu thiid-pei^oiL t>LJh^. fir |i|ural 



T*blo2-13] The afirtemeut Hjrrehpondence Ibr (?). 

Note thai 1b<- *i6rHM»meriC rnriefipohdetiCfl »!n^'in sHinantic pip Ed Warlpiri. Bfl- 
ciuai: the ^uuliATy hn5 onlj t^o pui^LltJiift iJL It Jur nominal a£Ti»[DiMt^ tls*^re w no 



apeemmjt witEi tin* third Fkrgament ^'hGtl iclcctcd hy thf predicator. T1i5e cnike* A 
iDipMi^itile to Ea> fioiuelluugljfeH-^ 'L t^e ^on frtra] liim' ^^-iLti^jt ?U|i|>ljln5 the overt 
pronOna, hjn^tifu 'ytju ^ (17) showa LhE] trinbL4ttLH>]i oftlltP 6*rilencE without thp OVJ?rt 
ptonouih; Lt inust he inlerprirtpd ivitli ft t|ii rd - penion direct otjixtT dtl* to i.\\^ lark 
[]rTEg]str;Lti'>n kn tb^ a^Elltoxy^ 

take-H0MP^ST-l9-^ 

■r may take him/ her/ it frum yon/ 



03 



Chapter 3 

Representation and Algorithm 



Thtfl nKtptei prcEentj ii complcLo d4.^tk|h-tjrijlL oftLe TcprFMentationa and d^rilhiru 

of the [Urti^r. Th^ £0^ ^f this preEentatiaa Ie to &bow ho^ l]]t> par^^r laaficllei 
bath JjE^ and Rxniid-c^rcl^ ph^nnnt^enA. Ta den^GitstratE this abclity, I "^ill shov^ the 
parser proccsjang the SAm[sLi: sclitta^cr ill {H (rejl^l^ lieie from Lbe in trod uctioii] 
an<! aonie of it& permuted coaeinK. Spcci fK^aU y, we v^|[| gn^ th^t the paj^r derivea 
cqi3]val4nt fryi^lfu^tJr Stru^tiirefii fmiti whi^zh fnquLvzlenL Eemantk in r^rpruU iLonK n^H 
be retrii^ved. 

[]) N^juiif^Ti^ id-?ttii-rin punln-m* hjnftj-tu karIL 

^1 anj lakJiLft th-> toooLtitiacig Irom ihe child-' 

Ordtrlnj phenomena do not conatitut^ tke only domain o( tlic ptTsCri JLOW<>vet'r 
A more nearly incLusive list of the Wailpjri ijliCJlOCtieiia thiLt are handled \s giTEii 
In lable 3,L Tb« du^'n^ion below will aJ£o demonEtrate how Ih^ pat^^r conipula 
each of these phenomena. 

precedence nominal, vfrhaln ftfld ati^iiliarj' Eorqpc^Etjoil 

continuous ca^e phnuL'fl 
auxtHary poaiLioninfi 

tjintr^f i^alnmatical functLclnff 

free |>hia±ic Otd^ 

KmiLnLicE argument idenlllic^tion 

null auxkiiary compon^iltA 
tense i^d IrgUiUCnt at^rt«meiLt 

T^bl? a.l^ PhcnomiMia, fiANdJjJ hy Llie pajtEr 

When reading the des4_rLplicinv bc'li^w JL Lg iniportiLnr ru Temember which sCrdi> 
lun^ aro r^aponalJul^ lor wliLch phenomeiiaL The pbcncrUlcna involving precedence 



ore prcHCfi^cii whth precerlence ELTUctarc (P2)i ^ht jynLaxiii pJienuEnsna anj han- 
dled by s>]Ltactic Slraflnre (SS)- Sn^cTiajltH^ pfOCPS^iftft Ifr fti7.eom pi whed w|tLi a teC of 
InCerprecLve rout]DC£ that operale oti SS. 

TEli^ nevt $i;^t|orL dl^cms^ th^' r^preaf^nt^tir^nR of t^Qbh PS and SS^ as weiU aff 
the l^coB. S^ctinT^ two pre£eiit& tbp ali^riibnuiH and i]cnLr?netr4itt'ft llndt ^h\t\Xy tn 
hlLnd.]^ L|h^ plh^ni^ni^riFk Lk^t^ above. 'J'he la^L fccUoia j^ves A tjac? oE parsing Lb^ 

ipLe E^Fitence. 



3 . 1 tte pre senta tion 

Tbe parser ^as designed lil il]l ohject-orHiTited fltyJe beCMI&e at seemB to capture the 
Sfirture of Cov^rnmertt-BiDdjDg bas^d procesEin^ Two in^jar rjbjnxt^ iti IhA p^Ri^ 
are prcfedei]tc*tnai^tu]^{FSt AfkCl fli'nt^tl<^ structufp (SS) The other major object 
la th^ te^ti^oi^, whhch iH the repo&ilory of inronnalioiL Tur L^a^ I^^XJCAl JtOfflr Fir^t I 
dL£ciiA£ the output Ebnictur^T *^ ^'^^'^ ^^^ lericQii. 

^.1.3 Prec<*Jciict Structure and Syntnclic Structure 

BoLhrSajid SS ircbasttton tTtW. C^fli node In a tree noqtiJDs a category UbeLiiid 
data and actions partLcolar to the Icvcl of rjjpttist- htittloci- Kor example, Ln PS there 
are ^ictJOaiS fyr ^Oiubinlilft Ad^frt noden Id ftS, on ihe olher handi *titicnl& toay 
not u&e precedence anformatLon bt:<aue<^ i4 h ]Ll>L r^pre^^nted at that level losteadp 
lhcr« art aclLOf^ fOT coml^LnaEtan of ayntzctir: strticLnrefl smzI\ ^ tiiA^tstniW^g ^nd 
f-Lsai^ment 

PS L5 aLtnally 4«i wdtrtd fcUKt Oforderwl tiPtg, F-^rh Irte rapresentB paits oE 
(h? I npnt sente^i^le where precedence is ^^Jevant^ siicbii artlMl|! thn^ inorph^m^^ of a 
Word. T]lc rtlatir^n butWH^n tf\^ tfj^es in Lhe fore^l h not relei'znl to pi^KeSfiiDK ^'^^ 
sei^tenne: bowever^ the ordering is kept Lo mtirir the or^ltr of the Input sentence. 
Bacaui^C pht^*^!^ ATf not <iTder^ ^Uh rc'^pect to one another^ Ih^^ PS for Warflpul 
Kntences t^lII not conlain tr^es i^ich t^o pbrlftc^ iiL thcm^ r^tther^ tb^re vill be one 
phraio [Mir two. 

SSt on the other bandi Is an nnordered Torc^t cif Utiorderenl tr^ca- only hierar- 
chy JS Te|>J^entod hr^re^ The i^e^^F Jot a forest rather than it ^liigU' tree is a Ihit 
Ubihtle. Followinft the fiB juLnciple of Fnll Intciprct4ttl03l— tho HquJrenient tbit ev- 
ery tlement of ayntittit tcmtturc JKOLVh^ atl IntorpreLatlon- we would expect tliit 
gtaininatical Stntoncos fj>j respond to a single structure in syntaXj thftt Is, thit no 
eleineDt be left nnat bached because- li iacl't Ud^U^cd. T'hl^ IS, Indeed, a condition of 
gj^nunatkal Lty, Ajid t|ie p.ir^er checbB this upor compLetion of tbo p&rao However, 
a Jore-Rt is required because durLng the parse thCJn^ m5.y besftv^ra] unconnected trew 
cort^pondlllg ti> different part£ Chf the input sentence. This is <t toy ti> the process- 
in^ of free order phenomena. Consider f2|s whitli la a variation of the main example 
qentnnce: jLb input representation is ^>eiL in (3]. 

ch|[tl-DAT IMF^HCtr-t- 3 l-EFMI bcwjin crailE 1 rillO- N OU T Ai^T 

'From the child 1 am tahiil^ the IjOOliicr^Eig,^ 



(a) ((tKUfirHJ KU) (Kfi RBA HIA)) ((WCAJULU RlU)) [[ICiJtLIlJ 
((PUKTfi Il!fI))J 

la Ihe proCKfl i>f parain^ tills icnlfaine, which is ptrRimn^ li?ft-to-rifihti the 
p^Tfi^r will reanilL *^tagt where il hftjj pfOf^Hie<l alL but IhelisS -A-orcl, j3Ejrt^fl-rrt(_ At 
LhiE poldt It will have pars^rl rlhf auxillaj-^- wcrd, Afl w^l| iR eacli of the Ihrcc tiJC 
phraseB. BhWAHS*^ the veib hi^ n^il y^t fjkt^red llie parw, Hlfr^ will be no way for 
llie afcLNtfuctuwa to bf j:j>nnflrtEd, anElciwI, llity inu?t ieRid€ Beparalcly, Ai: ^liownin 
fifiiire a.l.^ Wti4^ the verb dOoS eiH^r the syntaclic tltiirture^ tbo ajguments may 
tf roiiDectcd b^ jnsoTtlng them into tlLft4rsuin*Mt.poeitioci£ ^ Lh^^ V'^^h'^ piojection. 




Il^nju ^Li flu 




© 



Fi^nr^rl: r'hei>S after pacing loui" wwds of [5)^ 

OilH^oc.lLcrii]ffereiLrehrtweH^li PS and 5S coiic<]ilift jn^j^f r^JJ^n^ Eac h nioiphH:mi? In 
the input fitd (Cliff I& projecltrtl into PS aeftylnglp Dode. However, i]i SS ^yntflxtically 
nhvant parta of the anptit M^nt^nn-ft project ]nto L-*tructures, which cnjty contaid 
tfnwi one Of two k'nelE of proj^: liCfccl. L'5 tiUCt u rea axe encoded i]i SS wKh the Ftid of 
Ike ^prc^Jflf-tioTi:" riajr that Is Stor^^ with each nodv. This flae 1& trne if and only It 
itB pax^cLt Ik a lliemboj of its proJFclioHr 7hb reniaires, of courstn tJiat eVAftlj one 
of a cLCM3e^3 rKLldfen have a true projftrtlon flag. Thi& Wtll-forcUftdlLegFi cojidiLicMi is 
m^i by the conKtructior ofSh, is explaidJ?d ia KWtloii two_ 

An exacliple -should clarify tht Kpr^eJilation- The tyntactic ni^mifeatation of i 
Cftf» phnue cOilsiilfl o1 a lise-marker lliit liiW pirnjectpd one JeVel, lakllLg ttin^ JT-oii- 
tltllUClLtnounG as arRQn]eiLtfl. CoJI&tder the ciBe phrjit Id |4)r 'I'li^ LaM-marhcTj rlu^ 
hK thr^ argument noiiilSjlfiJTin;iiJirrorii^ and tttnTirriM, This ib shown gr*ptdtally 
in figure ^.2. 

[4( pirnnjF ffifram hMTdirrpa-riu 

tedltpede homesick hravn^-RftG 
'the br-tVO, ^om^il:k centiped<>^ 

Tbe fhirser'a rcpr«ir n Nttlo ri for t he Bame ph r^tii^ Js j^iveu In fiftun? J.;J. Th*- [nWl^r 
djftplri^tt its rPinlta on it& ELde. to ib^l the lop of the proj^rtiort appears to the left. 
NolB IbdJ the left-meet nodp> ^'H^lvg^^>■ i3 case, as tt ifi tlie lirstr|eH projection 
of tbp gaEH-marker, rJr^ ftlLi}'.tn at Lite bottom of tlw ^trncture. Note aJte 1lL^^t the 



AlxilS^ry ilrufLTiTEi arc ^Vbh bclil^^r 




jpmiiji ]prTfl."TJ 



Figure ^-2' 1'he gyntif-tJc m^Jilf-Ktatloti of [4)- 

prujirctton fla^ For the cafc-mar^^r Ls Lriie, ]i]djcEt]i]g th&t it is tli€ cliilcT of li^t-mcdt 
node that la In It* proj^tion; the olJifr liMos have projection Rtga tliat ate faJ*^?. 



SL 



ptoJ*ctton?: HIT- 

E^tagory: CJlSK 

children! prci^ection?^ HIL 

catPg[jry: KOUll 

pr^jJftCtiflo?^ Hit 
iBflrphtm*- VTWUFLtl 
Category: KDU}] 



prp)«ctiQn?: HTL 
QLorfheiDfl KAMiIFLEU^il 
Htsgory: KDUll 

prDjectionT: T 

£S data^ CA5E-H*RRFP: EH-HATIVE 
iDDrphsne FlLU 



Figore3_3:TheSSfor(l]_ 



3,1.2 Tht Lexicon 

'3'he Iexicofi l9 repr&Beiited i^iihi|^l> a^fl-P^L fpf palr^ of Ll^kIl:^ Itetna and enlriea.^ t&cb 



ft4t }l«ntJ l« Ihu phnioiticiiGD. bu. It in UHirnHJ L ]ial iz^di Jc^d^ iIveh tti^lM hllO^XU]llv ORC CA^- 

B7 



rij4htAiih only lliu LiiRuni^llyli LIlfiL is t>arLLrii[,tr Un ih^ IfXLcal [leJn. Ftar -^K^mplex 
proLioiins JT^Jliniri iheir penon and nih"mhfT ] nformatrati p but ILDUn^ do no! bixiiiflC 
[hey all baVFz Jn^lftlllt v:i|LqG£ €it lh]rd-p4^rfc:urL, ^in^uLtr.^ 

Figure ^.1 ahowfi tbc [cKlc^n U«ed by lL^ parRcr to |iarw the simple s<inti:fLfe, 
[I). Eacli loidtlJ iLpRi L9 gtven Ihe calGftory lindfr wtikh eI ]t llHtHid- Foj cKacnpL^^ 
frfli'fip iiEirdiJj and nffo^rJil ftl*^ ^1 declared to W ]30iini- Kach item lia4 aftsnciatwl 
with iL oplLtiCLftl Liiformatioii, h-^hich ]iia>' lie n^iLhei for FS or S^, ttifl tri1oTmatj€]ii Inr 
5S iH bolh Bj-ntActk anrf apinarlic Ln uatRiR-, Thti* IriformaUoii may C&m^ olther in 
the force! otdata or acUtrnB. 

A fe^ of tilt entrlft? an? ]iiRh[iEtiLj]d llCit^ tliP TSTnainder uf Uiii l^^Kon wLLL he 
dificniiiiod helnwln llie aectjoTi Oci il^orl tlkin^ The proi]oiiLK n^uufifii Lb diatinfliiishcd 
from otber ucmns btca^Lie H5 penor and number m lorm^iLion ia stored ih t ti C lexicon h 
II ne^Jr'l liave MlllTflreilt cal cgorj^ Ihou^ll. lnW^iifte nouns and prnMIOIt nft fiirt alike id 
botlL prwt^kt'Tirfi and syata-^ StructiireS-"^ Aiifiorialed w^UlL t^fh auxiliary elcim>Dl \e 
lllP nnfnber oEBjllahli-^ it n^onlains- I'tiia in forma ttOci W n-=:wl during ^ftord piifsineto 
check ttHdl-formftdjies^' each v-cnrd in Witrlpiri cnusLconEo^t oEit J^aiit two syLlahleH. 

MiJ<^h lexical ]nforinatl<fc]i appbtifl not to u &iiL^<^ Item but to entire cfasse^ of 
items- For -iX^L]rt]Jlt^ all verlhs in W^irliuri that ?:eltwt an aci^nt P-rolo ABSign <>rf^livB 
ffctr nS^nce LhiB ci£e-:Lbhijift]iK-nt if: a feature of all BUtJl vn^rhB.il wouldu't ho ippro- 
priato Uj store t|if^ action ]n J?Ach vjjrb'B entry; insteadi iL b fltated onMj as a rule. 
These rules are represenU^t st j-aJgSi t lorwajd [y iis atiit of pattern-action nil-a. AftW 
lejtical look-tip Is perlonncd, thp llit of iulftfl is apphed. If the pfttteru of Lhe mlc 
tualcJlw the category, th*- ru[i5 li™, ftnd the infomlittltm fcpwJfied in the "'icti™'' 

part of the rul^ i& ^d^d to Lkc noJe^ 

For e.iumple, couEidrr thft lesii^al nilea that encode the mtlllftstation oPff-erida 
in SS. The Hrst S^^t of rules, ahown iu tstblu 3-^, IndlcatcE the uuml>er of levels of 
projection for the L-ftttUnZturefl of certiiu cateEoriiS. 



LI the item tb ^ r^f-inarker 
then it praJi^ctR one leVcl. 

IT Iht Ltr^m Lr a verb 

tlien it prujccUi two levels. 

lable 32: Lexical ralas for [hroj«tLon. 

Thf^rieKt aet oFrolraioni-erns Llie IkcnsLnR of t^i^ie-J^ipni^ent actions. ThM If, 
lh-*e lilemftitR must be pEGB<?]it in tlie ^pLeuce for tho cSj^4«algnment action that 



^Aclaal]^, M ]niirlni||i ^l ill tfct lutrod ikI^wj- [:li:i|»t^t, ncilTls linvE a detnLt nUnlj^r of fltfcw 
KLiJjuJar Dr pluiiil ri>m h] n-^tjciQ wilt ll« a^iecinriiL ciKk* &f l[>^ Rnuliicy dt^Lriilfci liiai "^llLCll 

^InlfrtfllvufsicBuoflficpiTMr, ttiii^hfifllcLkiin iriy WKCLIiivetotieimtJflJrKrlfJinttccalEsndal 
UlfiJtltJ4tl[ip bnixLiw [iTUEU^nnj ran va^t 1 1 i4o flOBlt fljnticlic cauntrUrtiulLi tllit Pn"ni cannDi For 
iA*»[i|^ Ml Epe'J"!^ '^ " perfcrlli l(»[iF-ilh 1* h*« ^ bqhb pliruru- 11)nri«tin|; pf a pruuaun hy 
llM-1f, |i LJ n n^aEn niaUL-il hllliriy CO fia'^^'J bdhh. tiDLi^K Amfcl d tLf r li^ p^DrmJ oi kare a iJctTTittllltF 
ipptw Willi ^eh. 



Lnoun 

(HCAJULU 

CKU 

Cbb {data (caE€-A3 3i^ed dativa) 

(PUWTIi 

(pa (data IcoFiJugAtion ^))i 

(as (dati 1th&^a■^^l?lBB Cagent tti«DL4 p^^h)))}]) 

CBHT 

(pB (data {conjugation 2)t) 

(Kli 

(ps (data (a/]lph]*0 l») 
^44 (ditft (tflnsBB (n^npiBt)) 

(ajpftct IpptrJftctHHJ 
{ AUXil iary sub j b ct 

(pa (data (ByllabLea 1))) 

(iiujib4r aingular)])]) 
(au^cil iary-daTivG 
(HLJl 

(pe (data fByllablaa 1))>)1 

Fignr? 3.1] The porticn of Ltie 1e]ucon noctlcd for pat^itij ^entepce fl)_ 



H llcfiiues La TiPCPmP mnnUf-nL Id ific Btiutture (lg-t addoil Lo ihr SS wtiopa of the 
jirojpctlon), 'I'lio ruloF are gaven m 1a.ble :i 3_ 

thEui H llMllBiS SMJ^flunenl of er£ailvi» rts?. 
Tr Lbp ilEin L« ft l*TliW elomei]t 

If th*^ ttom u a dniUve iZfrS^-marker 
then it ai:ElKllJ> dai|%E cHae. 

Tablo 3.^^ LgklclI rul« [r>t' ^ftae-aBShKimiEDL 

The tfilrri aet ofmlM i:^ for Lh? Ltcensmf; d" fl- AASlgtl m? n t, as p^*^ 'n tahle 3.-1- 
Tliese julcfi alio elu^w the path ^-atftJpung pioperty cf datLi^C cA-ie-markeir Thrt 
ff-Tcje L9 combiinjcl iyrtaf LLcaJly ^aih tbH_- prflrfKAtor'a L-bLniclun: via l)ip indirecl 
ftLep of ^-linking wliicb opcrak* OJk ftodea that haVe bwn fl-MEigned. 

If ihf^ ilera bpI^icIe tlil aa^nt #-role 

tlltfl It llfpn^es fljifiipannjcit ol lll&t toLe. 

U llifl jt^m selects i theun: fl-Hilt 

Lhcil it Lk^nstfl aEaigmnenl oEth^Xt rolf- 

If tJL? |ti?m aelentB a path rt-tolo 

Llien Lt licuny-^ h tlklngof thai rale. 

If the htem ie a dative nz^^-maTlxr 

ihcti It a^&lgnFi tbie path fl-itlc- 

Table Jr-3; Lexical mlea tut f-ftfiekeiim^nt- 

Tbe aheve two tlhlts Rave rnlfis Larj;ely for Ui^nEinfi caEe-Efe^ignment and 0- 

afiBifihm<it] tactions- Theruln]* in LJh]e3,5 dlftaietn^eTTtliE^actln^Eaje tDhesltuAti^d 
In l]\i= L-BtruclnrL\ ih^fi mies det*rminE gmnin^atical functir^n. Wtien these tuIch 
IfllJ: ahaut bfjn^ iiiwiiJeeled as a tJ_^taln gr:tm mahkal [uaLCtiorLT it meajit that tha 
Sct4oil^ foT the CDrrespoadltlJ cafie-aRsiRD ment anc[ S- aftslgnmen t are fautlk pUj?^ at 
Lha^ l^vi^ IJl thf proJecLiocL 

An i^xample thmjld halp ti> riarcfy the Dprr^tion ol these rules. Consider the 
parsins of the Vftl>StH^m, punlo. [n tlic lifSt t:t4*p of IcMiral IouIi-UPt the item'B eatrj 
[b iCtrLcvwl from the le^iicon; t^^? entry is shown in llgur? 3-^- 

Ea the tftioild Step leEuia] niJe:^ are ;tpr'1lP-l to the entry. LiXnlElng to the fir^t 



If AH agent ff-ro|e ib selected 

tti4:n U will b? tn^ikiJctt ^^ tha iiibjc>cC. 

tb^fi Li ^V-lU be itiiiJiifL'ftt ^^ ^Jl object. 

U a tliEmi] f-rolG lA fr^k'^ r^r] ^nrL there ih did ^uhjecl 
then II wiLI be m^nlff^l as tEif ifuhjAct. 

If ft path ^-role Ih flnj^tH 

then il will he inMiJf-esi as ^n t>bj«t. 

Table 3.5: LGxica] rules for deieimiiiLii^ ^THmmBlLral fancliDii. 

CPUST* 

(pa Cdita [conjugition 2) ) t 



Figure 3.S. Tbe lexical enli^ ioz piinio. 

aet of rulea ii3 Lable IT^, wa Rnd one ruJe Lbal appLes~ i^erbs project two levels Ln 

syiiU:{. In tho dh^I ^tt of rulee for cigc tboK If ono rule th^t ftppUeSr Lcioking tQ 
the verb\=: ff-KTlrf. w^ fiee thfl-t It i3fM»g e^^l^^t fLU ftg^nt tf-roU; accordins to the mi^ 
Kh^ V^rb tlierebrc bceiuj^ AfEl^ineT^ r ot^r^tlVG case. 

In Ihe E4t al rulefl concerning; ^-3£H]Rninent three ruLe±] apply. /^unCa ^electc 
Ail ftg4^^ A Ibn^nlH^^ And a paLL; tb4^r?[or4>^ l^cDrdiilg to tbn.^^^ tihl^^n it licerid^ S' 
aE^i^nment of all tbree More precEsely, Lt hceiiEeE H^Ei^ment of ibe iirst two^ The 
path S-Tole IS MtuAllj- Mfljgncfl by ttiod&tiiHC <^ft5e- mark w; the "verb indirectly asEi^E 
Ibe f-role via 0-hnklng. 

The last set of rules detertnines the graminathcal function of eacb of Ihe aigu- 
mcill^r FoUovfmg Ihe ftiit rale of Ihe scl wctcc tbil Ihe ^Cfit fl-rc^le will appear in 
thr iuhji^t pOfitiun. Thai mnMnt thit r^r jjitfiagnnbent irf htn ewe, erftaCive, ifld U- 
a5sign[[L?nt of ir^ rukn a^nt^ will taL? place in Ihe specifier position The other two 
tf- roles wUl appear as cbjei t&, ao tl^«r ca^e- and #-aEsi^ merit aclaonf wiQ b? pla<rd 
in the serci-le^'iJ pro>^t|i>n oEthe verh, afi- i^ilillng tjlthe complement pcsitiont. Tbc 
computed «nti> Ls ^hown in 1ip;ure J.ti^^ 

The Lnter^xtlon between Uee^isingr case and tf-role operalions ih best illtUit f^t^n] 
by nn »^s:unplc- Ftgtire J.T nbown th<i JilOf phoh^ir^ paie^^ for phnta. The PS compo- 
□ent a( the ontpnt ee rather ati^^htfarwa^df so I con Linue with S5. Fjfft rotlce rh^it 



*TliH ^mUy FlHii ilthrf hf ■■rtiftjuiLed U tin IrKhZfc] jt^ ni IV fluir«il into ihc Iciii^m, ht mi*n¥\\ 

}aoV-ap of Ibr it^ipa flihn^lfl tl|4- ih^fAf* TllC l^^T^rrH ITl bf L, Cakcx ibe Curmer lack La urdt^ tO M^iT 
bmc duTin^ r^ii^f^ 

71 



(ptnrrn 

{ps {data ( coaj ugaticm 2>]) 
(bb Idata (th*ta-TQlaE (agent th^me path})] 
(tCtioHB ^prnjactiona 2) 

(llCflna* (cBflfr-aaeign . &r£atiYeJ) 

(Li^bn^fl (theta-assi^ . th^mft)) 
(licenBfl (tbi^ta-LinJt - p4iKJ) 
(jp^cSriar (iaaa-aaai^ . ar^atiTo)) 
(ccimpiaaent IcaGB-ftSSlgn - abanLutivaJJ 
(c^nplehflftt jtl^lltB'aesl£n _ agant)) 
(co^laaent 4th#ta-a?[=L£n - ^hai&a)} 
(tcBplnh4nt <tft*ta-lin]c path)) 

FLj^re ^6: Thu comEmltMi entry tor ifur^ia- 

jiUPifl h»i proJFwLed r,wo lev^lss *^cc rjril i mR lo lln- spft?! Ri? a^tlOD in itg projactiona 
action. Oncu projuctrdi tb^ S^ pi^uar places pu^jI^'e ^yalactic acUoiiS In ]U StniC' 
tUT^h 7'he actiDOf flpectfi&r ^kI i^Bpltih^l^t pSafC tlics* aaqgnmenl aciioiiB at 
the fint-[evel and acm-l^^el rK^rff* fcJl il^p? verbis projaciioDT reapaclavely. In itnlflf 
foi flpaniiiar am] compiania nt to t-^ujintc tlnirr inLWt h? ^ i^^rrpfi pond ing license 
action in ILg HEru^Tiir^- Puwij \t^t lircn^^ ^1 buE cm? ^cthon^ nJLn)^>|^ \h^ tS[^^ 
aEsignmaDt actton for ibfitiLntivc fA^e, iithlcll P arLually Liceiued by tens€ element!, 
Ab i tKUlL this iompLament actLon muBt i^alL ihe arrival of Kh^ t^t\S^ ^orrn^Qt in 
ord€r r-i:>r [t to ilTV ^]ir| pWa tlia c^sa-usL^ciLng action in tLe Blrucliln>. All 4>f ||io 
Chtb^ ptacamani a^licmi: haia llrcds hnj^'n^ver 



3.2 Als'ititlim 

An overi'Lew o( Ibt parttr's <^^r*MH>T» Lr given id JiRnra,1.S. Input ffccitErnCHjs ctJ^glvon 
tfhtlie Pfi paiwr that traVJ^rafti tliHlS l^fL-tO- rlgli t and huildaupPS. EvaiJ" time a Unit 
of PS lHztomi^&yiilaH:t.n:All> rejevart a£ datarmLnwJ bj" thtltvm'^ U'.xi^ entiy— the 
unit k Bent to tha tynUu.tit panox- Tiie ByiiLactic parser accepth the inciMoinft biliti 
proJHJtt* it acturdlug to its laxical inrorrnalions inJ th^JI ftJItfln ifc inlo SS- When 
linishefl, Lhc syntFiirUc piiFEfrT roturnfi roncrol to the PS parear whlJiln^ouJUitit* s^mw 
n]on> iftpLitr Upon f^n^pletinB ^^^ Jnpnl HeutB]ic<?f tlkf PS pRTMT alopa, and bnfii 
crtiLput structncTS ai^ R-tunirtd- 

F'vwEi Idificnss thr P^ [n^x^ri and demoD&trata ]^ oparation ^'itti OKAtnpLc^ titivC 
rowr tll^ tanjt^ qT plianom^^na in^^ilvL]tE prtCadeuce. Tlltfl the synta^tir paner ifi 
diaciih&cd, 4^^n witli ri'xan1pll'-1^hov^'Lng its ability tohandfa theadvartlfladflyntactkc 
pliDnomend^ Tliis sactiOn tdnfLutlf. T.vJth h prftfieutailon oF the BricanLic mtciprtUt- 
tivU thftt il)e p*T&*r piitronufi on llhe output linLiuitic Btrudtiltt, 

7a 



Efi 



[}; data: CDHJlTGATJgr 



pro] BctiDQ^ : If IL 
C*tfrj5ry' VERB 
ctdldTfln: projection?; T 

RctitriBi CASE-A5STC[fr EMATIVE 

THETA-JiSEICN: ACENT 
catBgory: VERS 
cmidrAn^ pr^J«crticiD7[ T 

«liodi: THEfA-UNK: PATH 

-raETx-ASsrcH: theke 

ConPLSHEhT^ (CilSE-JiSSICl . IlBSOLITZVE) 

data: THETi-ElDLES: lACENT THEHE PATHJ 

IGEEfT. ERG ATI VE 

THEHt:- iBSDlimVE 

PiTKL DATIVE 

SUBJECT. .lafiErT 

OBJECT- PITH 
iDDrplieniB. PUNTA 



riglirt 3.7: Thf P4 and SS Tor punfo. 















hynlBCllt^lj 

4 . J 










inpni 

&BlltBIir^ 




PS 

par^EiT 


rrjf^vant ELiuta 


svnLacLiE 
parser 






























FS 




ss 





Fte&R3 3rB! Aftovflrvl«W"f ihBpaTwr'Bflnw^if rontroL 



73 



3-2h1 Tilt PS Patter 

la t!ifl Hret |iiirt c^f thi* ftHvLimi I diRCJBs ilht> alj^rLllun^ of iho |^'^ p.LMrr. K-ollowLng 

The Buk Engine 

Tbe P^ pai^EriE a recursive EngLnE Ui^t Li]j4jr<tt{? nil ih^ fotlr |7]kiiilr>kigk^ THzvelE of 
Eh? ilipdt i^lltej^4^^- Th^ tfhp kvol ol the piiT-RfT acceptji Lhs ciitLre Fwnt^iif^ ^ Inpn^- 
IL c^Ie on the phrasal pai^er tu p^j?^ 4>ach of thnj CUil^tltuciLt pfiru^s, and Lhcn 
I^OfJorm^ :^]Ltoiitial Ajztioiu on iho TEturnod phiBsal ELnicLursa. In & liLe mmni^r, 
Lbe phroE^ parser rails on Ih? ^urd-lEv^l [Mr^Er t[> par^t lOilS tLllK^n t ^urdii. The 
wond^l^vf^l pfLi^f^r raLlE on Lhs morpholo^LE^ parEer whLrh Is esBeDtiolly the laok-up 
rotjtjji€ [or tiiE kxjcoiL. Tlila Is diAtntnnnxl Lfi li^ure 3rS, 



wntence 



X 




sirnlE:]LtE^ 
parEsr 



phn£c 






4ard 



WOTll 

pAjcar 



nif^rplieitte 



lexicon 

lOOkrllJ] 



PSCrwphiaK} 



I"S[IbrwQrd) 



PS [for morphemj;) 



F]f nre 3.?: Tlir rpcurdon d( the PS parBer. 



Elsnh IsveI of F5 pajxiii^ bsee ibe Fame EiiL^nE?^ The b^^k alf^Drjtbn] la givei in 
Ld-L^Je J.6. T^ -fir^t «t?j> ol lilt ^gotlllhi]] i» ft It?ut> ^^ftt tr^Vnzniii^ tL^ hJhpuL TrDiD 



left CD TL[;bt. F^eIi anct (rnorphoin^s words or phra^?) Ls Eeni tc the tQbcird]TiAl^ 
I^AfH^r S^f proi^O$Bm^ Th^ &tTu^ture r^tLim^tL ja tbeci ^d4^ to end cf ihe PS lor tbo 
cuttGiLt levfil. At this [n>]jit uncjLwuti^ FS Mtlone an; cried toj^w if th^y Jraii ^pply. 
To aid in tbe ^ITLciency of processmfis tb« parEGr€inplQ>s ao^ aujcJUary stracLure^ the 
i^t of U[isatLellH}d ptH^Iltattti, that cotitajn* every norfo iii PS th^t lirw at ]<safit one 
unexecuted B4iLioaL TIiuj^ step {b) of ihe mzin loop conKJsU uf a traVn^raaL uf thu 
i^t, itteniptifLg to excfuti^ ^^h oE Ukc atti-cpn* oi the iicmIos wtthlrir K^rfe t]iaL only 
the aclioiLS that pertain to the rnrrent levsl oE parbin^ are i^onsid^rEd. 



l- Loop through coTiHLitueDtf from left to rij^ht: 

a. coll the eub ord inaLe parbcr. thi^n 

b. execute ^pplifabSe ^ctioiui on adjacent tTees 

?. Execute defank actions. 
3, Cbeck wdl-forrnedncfls. 

Tabid 3.6^ Thb VS pam^g enftloe. 

P^ antions are canstroined to aperaLe only an adjacent tiEev ia the forest. That 
13^ the actions in a node of one tree m^y onlj Act uil an ^juciit tfn^^. MOJ^ E[Hcal^ 
iniilly, beL^u^e Warlpirk |^ i^ bead-final lan^ia^ PS actions may only apply Lq tLe 
preced]nj^ tree. AnxiLiary procc££]np;, ho^'ever^ \s a special cl:^. Th^ ^tlona Er^t 
iiLclJwy r^omp4j4iiloii optraie wi tJie EU-^cwdJng tree, rather than tbe preceding 
Doe 

ThelLst oiPS actions is ^vea in table 11.7. TbehT^t routine concerns tbeiatetLttC 
ttjthH^fiyn tactic prui^SHOT- Moflt le^r^al neitiu arp relevant aL ihe morphoLofiica! level" 
Eor them prnJect-into-2S will be a morp boloj^^cal action. WLen ^hiif action tir^lp 
the syntActn^ information ^Jtlie lewcal itejn w jyven to the syntactic parser which 
projects it according to ir^ lexjcaL inlbrmatDon and ent^rR it iatoSSr A link k>eLu'een 
the i*S and ayntacLic nodes ia Lept for future processing. 

The ^normal" actions are nsed by tho non-a OAjliaiy ekmcnt^r S4l4Ct [^ tbe 
b[Lsic opFraCkon lb at niaufiefi tbe selector to become a sLbting of its object (ths trQ4 to 
its kftj i]L F5, at tli^-Kinii^ ticne, tti*-KyiUM;t ic^OunU-rpaft of tb^ohjfi^t tR declared to 
be an ar^ment of the syntactic connterpaii of the sekctor. TbJS Js tln^ ni^bAlhl^ltl 
by vjhich ilLrectcd Ai^mnea^t jd-ii]tilLi:atw]h is pcrfonutri. S^l^ct* ia like sfllact 
except that it is not deleted upon e>LecijLion Tbja action is used to piiTse cudtLnuobS 
caiie phT^isch. The liUit rtctifm, injectn ip rilpd similar to J4l4Ct, hut tli< fi^Titantic 
effect djtfers. FUther than de<^lajJl3S theobif^rt to be«a argnment of the injector, th^ 
syntactic inronnation in thn.' ii]|?ctor ]» added to rtni^bapkioi-icijly apeaJckng^ iBJcut^l 
into) the syntactic connlerparl o( the ybjGH^t. Tliis *tti01J ifl used Ijotb by nnmber- 
mflrJcra and fj?naF^ eln^innnlH Jor feature pcrroUtion. 

Thr' JLi tions rrjr pi ' tf hi^^i ii ^ Llhf auKali^r^ woid hLfc n^nit^' lit'' KW rnth^'r artli>ns. 
^J^t-adja^^nc is u^H tXh build up thf pnecedeiice struct uxe of th^^ ^UKlli^ty word. 

7S 



ii3t^rfafQ to Sd projact-lato-SE 

normal acljoaib ^^Iep^c. ^^iitfcl^rJ) 
■•L*ct* (cfltfigoryl 

tight-idJaci*rLt (aiiziliaTy catfigi?riBS> 
aoxilj.arr-BaIe^t (aUJSlli^ry HiaiegQxy) 
4^XJliarf-i-aject (auxiliary n^Btftg^iy) 

ThE uREimEiil qE tiLJu LtcliO]M8fL||norLhecateflm-L€softheaiJ3dliaryjdertL^iltftwhlrh 
may appear to tliG risht of tbn3 notltr For H^JHiftirtp]^, holh auKiliii^ obj«L clilii_£ ajid 
daliv^ thlk^ jnay appf-ar ta tbe ri^bt of ibc babji9Cl rlM^^ S^ ^^^i\ subject cLhtlc 
wotiLd contain Ihc In^xii^al action In {&). 

(5) rijht-adjacsni; ((ftujillarj-objectn auxiliary-datiVflJ) 

Tho tw& AiMl^^m, aoiillaTy-Baleci: arnj autilltrj-injact are aualoj^na !□ 
select and injet^t. TLi-> (lUT-ererLe i^ llial these act]Dni: drj tiot build i^p \'S ab 
ilti l]i^jr coiinterparta- ratteif tliey fljiJ thtlf *.rpim^Jitfl in Lhe set of fiibUngE aA 
coahtrUctHiJ !>>' riftbt- ad j acant- Th* au_Kiliikry bawr &<:lett5 hs nomLral agreemenL 
cliLicR, finT GEamplEL 

The la*t au>riUit'y a<^t.|fhn> auiiliary-adjacant, ts osud for a-UxlHary clJLhiiii- 
LiatiL ThE usual adjo^'cl^y routine i-arj not bs UBcd due t« ihe BLmn^ n^ttirc of 
t]l& UiXilJ&ry^ in-itead^ this sp«]al routine ifr icS^^cL. auEilJ-aryadjacBnt may Eire 
U thpre ia EomElhinfi to iLs lelt, bui ai^j structure in huilt up. The purpose ol tbilj 
r^UtiJlf^ is to ^d In chenLiuR the well-fomscdntti ofiUJilLlary rlUtrL^aLLon- 

'I'hft diftcnaaioD ol aoxlllarj paramft Is L>H>st rounderf cuL with an example. Cdfl- 
£ider ths auxiliary word In (6), taben from the BampiE FETiU:l]tHi. The wmputpd 
entriaa for each element the ^hrj^n in Hf^Life ^-li- 
fe) (KL Ml KLk') 

Wsfitttj-L wltb tbelExicil parfler. Foltcr^'lngiitepoTae of theal^orithttin thni paraer 
rails its ill hotd initio, the morphaioftical paiMrn wltEi TIll^ lir^t ]norphemfi, ta ThikL 
p&rfier Hzopipntefl ita lexical euLr^n £u shown ftbovni, and ureateB a node in PS for it. 
ThesKOnJ pttrt of the main loop is thsn nsitbtd. Thw^ ]& only one morphological 
aj^Lion, namely, project-intfl-ESr Thfcs M^tinn ralh theayntictic prDct^sor ivlth the 
atlJTlliary nodeasanarRumenL. Follohvilig tlk-prE^jflCtJona artioa, the node projerta 
tivo levels ill tISr "The flpecifier action Etloo Hirw Ij^au^f thi^r? ifl a correa poud Irg 
Ilc^nBb aetioD in the EtrUtrorf. At thlh jio[jit no rTiorea>ntactic action* rOrtLiln, SO 

ra 



(pB (d^tt UyLlablflB 1]) 

(m&rphEjl o^i caL < pro j ect'-intc- SS J J 
(Itllcil Crijht-ad]acfiiii. C4Lqxlliary-sub]BGt 

aiiiiliarr-obj att 
4iUXll 1 4rr-dat iT«) ] 

( auxil 1 try ~fl fllett ^luzilxary- sabj *ct ) 
laun-lxai-j-Hfllact -^lui 11 1 iry* abj Bctl 
(BDirtll ftrr^fefilect -Buiillar f-dai: t vb) 3 } ) 

(aEpact lcap*rr*ct]> 

(11 0144 (argument . vaxbj) 
UpecLflar (ar^iyniTit , Terb))])l 

[p4 (dlta C«}^Ilables 1>> 
[action* 

I m^rphol (J jical (prcp J 4Ct~ into-SS) J 
(1 AXiC Al C auKi L iary-ad J a^aht ) 

( right -Ad J iCftCit ( aujil iaxy-ob j #c t 

auf i L i Ary-dati Va> > J } } 
(b^ (dAta Cparson 1) 

'hUaber Bingular]))) 

(pE Idata (sFllablfis 1)> 

(actions 1niorph&l<i£i.tal ( pro j a ct-i iito- SS ) > 
(Iffiical CauiiJlsry-odJttBnt)))^) 

Figure 3 ID" The COttlpntHJ entne^ for tSe phmentF of tu- rfW-f/t 



Tf 



E-dDLrcMs r^Lnmed to thj^PS L>Ji.Pf*r_ Ajiit tiirn^Oitt, thi?ri?arin nomDre tnorpfLr>lok^l<::il 
aLtiociEn HO the ntorplLDloRM a! ]l&r=:^' ol id IB com[Ntclr. Tll^ P^ conaijucled fi> far L>i 
rtturnoJ Id Lhi.^ ^]>:l^Jd piLiscrnncl atldwl tH> Ubf lexical PS. The output struct mea it 
tLiH jjtnrtt 111 the pane afft tllsplayi^ iri fignre i^.ll. 



Oi Lencal uttlOM^ ALTILLIAFLT- SELECT^ ALJIILI if: Y- DATIVE 

AyitLll-flV- SELECT: AUJilL I liBT- OBJECT 
At;l[LIJ.ilY-SEXECT: AcnilLIAFLT-SUBJECT 
RI G BT- AO J AOENT : ClUSILI AP Y-SUE J ECT 
AUX][,I Alt Y- DEJECT 



lUU LI ilItY- DATIVE) 



d^ta- SYLLABLES: 1 

irrarpliftufi^ KA 

catB^orj; AmiLlAilY-fllSE 



^ 



projBCtUn?: MIL 
category; ALT! I U ARY-BA3E 
cbiliiran[ prsJflCtJoD?^ T 

actional AUCIWSNT^ VERB 
catagory: AUirLIABT-BliSE 
chLldxeo^ pri3]^ction?j T 

data: TENSES: IMDlfPXST) 
ASPECT. rilPERFECT 
mDrphsne. RA 
cftCfg^ry: AUIILIAn¥"BASE 

Figurt: 3rll: Thf PS aiad SS afttr hftvlnj paTsed ta. 

Thfl J^^cal pai^r than liKWUtK the second step t*f tlie EiiaJn loop. Thf only 
node with ii^tlona Is the node For itft. However, non^ of jI^ in^tionfi TTiiy axccnte. U 
theitift lloaiRtiiiiPiil((>.Hliakc^tlploilBrL^L}in PS,Sc>, Lite Ltmp is [IcmU'^i, MliJ the 
iTiDrphoLogLtEJ ptiHerifi called Eor tlLHjflW<Hlii morpheratH rrW. It]-? parsed similaHy to 
ltff^ M]d a ser^od node uliimattly n^ntera tbe Lnxicil PS- Wfier the secQcid pArt oEthe 
main loop lA rtacliwl ftgajn^ there u itnicc^i s^]l action ihii mayf^fcnte- the auxiliary 
biiEc is s^fimbbeit wLlh the biitijn^t dttic ibronjji frinft t^t^ rLght-adjacent attioci- 
Ab a result of the prevEoua octioii, tlir ?mxlliaTy-Bal#ct n<:tlcitt RhJ 1 ht h u bjM t cli tic- 
may akofin]. TliiS tanflW tlieBiibjwt clilit to lin^flme an argijnir]ftt of theansiliaiy 
bue in 4 ff?niplement positioiU U^„ ^ a Kiblins oF thf ^*n>ti*veL proji?i:t!0[L. The 
two BiniclurtS at thia poLCLt in fJie p^rtn ^^f fihown in faguir J,IJ. 

A kev potnL hcit h fiow the IjrciT tLnnplltti^ i>[ thf a^uzcLliary moiphomea la nom- 
putcid. Thlii la effected through tht^ rJ-gtt- adjacent attLM- Its argnmenl cotiftti- 



0: catfljory- AfJlILURY-BASE 

childrfln^ 0. Lflilcil a«ioM: ^UII LI AHy-SELEfTT lUlILI UT- DiTTVE 

jLTintlLriHT-EELECT^ lUlILri,RV-OBJECT 
dati^ ETLLIiBLESr 1 
nrcirphfliM' Ki 
catfieoTy: JLUIILTAnV-BASE 

i: Xfiiical acTiDoa- RICKT-iDJiCEKT: < fiUI I LUlRV' □ BJECT 

lull LI ABY- DATIVE) 
datflj SYLLiSLEE: 1 

catagor?: iTDLI LTAflY-SUB JECT 

praj action?^ NIL 
«t*gi?ry^ 4ui;:liaky-base 

children! prQj action? : T 

■ctiant- AKGTmEBT- VERB 
cat»tory: AtJirLI ART-B4SE 
children' projection"': MIL 
datar PERSOKr 1 

KUHQEFL: SIEfCULAA 

cetepry] lUIILlARY- SUB JECT 

pro j action^: T 
d«i: n»5E$ <NOtfPAST) 
ISPECT: IHPEftFECT 

tanflgarj^ *UXILLA!IY-BASE 
Flgum i,n, Tht PS md bS fi>r i-n-rna. 



TO 



tutc* ^ clltiiitnction of the pobKiblLHio* for the siirrpfding pIgjdeijiL ed tliC: fctrmft- 1'ho 
dIajnnttLon allffc'A's f^'h rnmf>oi]ent ic he uplso'Li]- 

Tin^j^Lkier ki.^v to pronaiKin^ tln^fl-UXallAiy ccmceriu syiLtittLi' i^mcwalrg- Syntic- 
ticallyt the hf%»\l Ift HU^n&jdorfid tbe t€aJ of LLt ataXJlJary ph™^, lakiiiR tie a^etneot 
dltlK a* aJSimenta. This is erTcirli^ hy the Ru^lliary-a&lect ri^:tlf>tl. A-UXlltary 
bae^ coDtajn oil* flUCh acLitjn Ibr ehcK of the cEilli^, *nii every clitic chal docs ip- 
pt*r lh the inpuL sLrliLg I* t^^i^r M itii aijgument ol Lhj_- laSc. t^^ta that this Ibni! 
of BelctUnjn can riH^t hn folded into llic ftd>ij;eiicy action ab with ntm-a^ailiTtJj' com- 
popeiuls bj?taiihH* o]LEi eletm^nl JtfffMith LIle uljac^^ncy ^hii^ another (the base) ^RVcU 
the synl^CiC selen^ttrin- 

Thplafil raorphemfOfthelnpnt wordi rla^ ispalwd ju-^t Jjkf th? oLlteis. Oncy lU 
morpholoRkifc] P\i GfttfiH Llif |pwcal PS^ tl ih itilule adj*.i:eill in the auxiliary tfOni via 
Iht^ ri^ht- adjacent action of rrtftn and Sf^lwt™! b> Ihc haic, to, Tho fjn^ output 
StrtitHires for LhiB auxiliary wor*! an^ 3lbCi¥.'o in fip^urt: ,'S.I;*. 

RelTjnL]]!^ lo the PS parf-ing aljv>rtllini , t-Vh^ come Lo tike aecortd Bte]]: (ir^itorm 
default acLLODE. Ciarreilliy iJ^ere la oiJy oii€ denult PH action LhaL insGris the 
phoJlt>Joslc4l|y null abfiolulivt ctw^-niarher- After the phrasal pwr^f is norriplelEj 
the default acliuii chcrkft to aee tf the pbffa£Hj cndfl with a rioun that hac nOl l>Wrt 
IflHected Tor caae.*^ [n tiiifl ev^nt, a node lor tb^ abftdlutive ca^f-marker i£ inBCTted 
after Ul^? ia^t noun, and the PS |Jir«tr L^ callwl once a^ain to exeinte ttiO an^tloRfi of 
the ne*lv iuitftftvd node. CooEider Ihc abs-cJiatlve faae phrase in {1}. 

(7) ((YIJ4RIKJI) (fIRftABU> (KiHDIIUtPi) ) 

After tlie main phraf^l loop La*: rompbtPtiT the default aJition will td^tect that tbe 
LabL wonJ^tonfFrrTiw, hr^S ni>t bf^^ InHecLed for case. Ai:j:o nl I n^y, H will iiuert a nod e 
Tor the ^hsoluLivt caM-marker, ^'hcji^e iHiMJpTal mformalion ia sIio-aT] ifl ligMT^ ^-H-^ 

The default action e>LJ?cUteg tde nn^rplkologlcal and le?:ii;il ft^tlonn rilrertlyj and 
tbefl Jttuvokeg the PS pajBcr to pioccSS the m^xllfled phraae_ The nLCHrphulOE^liTftl 
lu:tfi>n c^iiBea tbe caae-Dkarkrr to pruject inr^ SS^ and \\\€ leikj! iJltion HlAU&ea the 
caae-mark^r lo turlevt thf Fijht-moftt houRh jubt ja uvr^rt f *ie- rnarkore would. Tbt 
pbrat^il ctftlOIS. a*lact+T Bclects Ihc pn!i:odlng iNOUniip again^ juat as overt tlflS- 
markerfi would. Flfiuic 1.15 sbow^ the rpRiilta_ 

The laBt opF^mllon of th^^ pupno ia lo chei-k pan*o v^elhformedjie£s* Tli^ natUfj* 
of ttie check depends on the phonolojical level At the morphological level, aa one 
miglit tSIJectn there are no conditioiii:. Tlie t^niWCllnft levelB^ howpver^ do eK^milna 
thp precedence strUctilttin 

At the Ic-Wf^ In^^'fl, three check', are £>erfOTmed. The firat makes hUCc tbal thn 
PS coualfiLB of onr th*e, (TinurNkg thit every elemont in tin? hfofd H^aji be aJ joined In 



"T^il« hCtfi^r> t^fci^ i~Lru ii> j^df? Lk? ahuLucy woid wbhcft niAy be liie lost af tllr pikriW, tf 

fiKu^lltq lh<: i^i^Ljni9-l4> Is^L word i\ ckfiLi ■]. 

^TJie iiwal atiTT oje-nmlSF JM JiOt ttOi«i hn IflC t^Tircm per tc. InatciiJ. lb** ai>d? CXJBtfl 3-^ L3|e 
^iJiK- nf :i Epr^aJ ranabln- tint Liii l^tfiti n-l MHiUlLam Ikr d*li *ril irLirn^ *>f tbe caM-rt*Tkci. 

^\\M t|i««H^ttLoJa[i crKir. tkc pirrcr minlf JlUcIj" li^l^J mil Trlnra-i *ii cinir mriMgc iJnip with 
h£> bfl UlhJ. 

90 



iL diTA STLUHLE3: 1 

DitflBOTy. J^UXILLARl" -SUBJECT 

TBDTphBiia' ALA 

Citttft^Ty: AUmiliRV-BiSE 

cbildroni pr€]JBr;T:lcin?^ T 

ac^Lo^LE: JlEWUHEJIT^ VERB 
catQ^cnrji AUIlLJiAY-HA5E 
Cbl^dxaa; projacticn?: MIL 
morphuLa : BLA 
catfifiPET^ AiraiLliRY-DATIVE 

pToJBcli&TL^: BIL 
dam. FERSDN. I 

irunBEEl: SiNGULJiA 

norphepi*' ftNA 

tal^giirT. AiniLIiRY-SUBJECT 

proJPCtionT^ T 
jBta: TEKSES: IHOHPASTl 
ASPECT. IKPERfECT 

narphema : KA 

catefioryj iUJ I L] ART-BASE 

Flfture H.ia; TIhe PS ind SS Tur ta-TTia-r/fl. 



HI 



(l^ij-cal iB^L^ct noun)) 
Cphr^fial jaelect* noun)))} 

(ss (d^ta |GA3e-3Lart3d QbBolutiT^E) 
dcti^ii£ (projattLoiu 1 tl) 

FigiiTH^ ;], 14: Tha [irxifAJ iidurn^nljoa Eor Itm abucLutivecaiie- marker. 

aoine other paxbL There ]b one eiception tbFit allowf the PCCGnd, UT^connertFcl word 
lo he Ail AuxiFkry word. Ah trusiiuorusd ;^^]l>v^_^ aLixJD^k-s *re not tntflrecl lulo FS 
LikeoihEr wgk^£^ so Lliis case muEL hdJlowedL Anjjliary w^ili-fDrrned hees l^ Eovurnd 

Tlie £f4ioqd cbeck eoEcres tliAt tbe vord (and the cptLonal Ejxili&ry wcph]) fcni' 
Ie^iu: at LiM^t t^rj syLLl^bl^r Itl tattp tins zilJ^iL ni^d ilr^t he n^pUclt for nrvuns and 
Vorhfl. In WarJpifE tberc arc i]o fciin^d-^yllablE no^iLS, eo aLI nouns^ bowevET Lbey are 
Innecledn will jwife, Verhs must siLw^ys h-e ijifl-yrtn^d Sw teais^:, aiiJ tlnco there ars no 
null i^m£E elemoutd, they Ioq will always [>aEa LLlh Last. Hence ihs check Teinain^ far 
:i-ii^lliw^ wnrdsj wh*;iv ilie 5y!lablf* i>[*MJ^h flhtnert ire sjmttifld and toirtpafcd lo 
LwOl Thifl ]£ why there ie no explicit ayllablF information for pon-^uxlll^uy lexiciJ 
Jtonu. 

TliE ] Fut lexicaF chcrk TTiakc^ EUTC that all ulitic^ ^te. infant, ^nrlltJC tos<tmethtnEr 
Thia IE ][nplEinEnLed h^ examininj^ the lexical actions for the left-inoBt node of tbe 
wont. If jt coDtain&an unExecutEd sfrlect or inject ncLlfhn, Lli^n It 1^ ^n LI n^atl^UccI 
tlitics and lla^nj^ a± huch^ This check ce eJso pErfonned For the aujuljaTy word' if it 
containf: an au j tli ary- ad J ac ent aHiEioPr thf wi^rd is dfM7|&ri?d ungT^mm&tical Afld 
an error L^ signalled. 

At thp phiasal Ievp] odIv one rhect In purilnniwl. Li^.^ th<> If^ral leveln phruea 
ars r^nln-^l to ^onfii^t njf onE tr^c {a^n^ with the pueflibLi^ rjcception of a trailing 
anxiliarj' word) This condition Jitema fpwn tk fajrt chat phonolopr^it phr^ee m^y 
noL contain more than one "syntactic phraae. 

At tbE Eentential IeveL auxiliajy poaillonin;^ Er I'becked The i^t i^ ^jmpLs' th^ 
wislliftry word. If prfspnt, niiieL be e|thi*r In cJit^ first or serwi^l pwlilor- OEtoitrMp 
this not the Entire conditEoo^ aa some auxihariEa are required to bE in the aocor^d 
po*itloftr if wt this retjuirement ja taken taw hJ ly (heclltlHTJaitjon dn?H^fc <nily ttL«j? 
aiChiLary worda that mnsC hE Enclitic Lo EOjnetbLng aj~e rcquirEd to he ifi &0^nd 
poflltLon. 

'IhJa comptetp^ di-^uRsioi of the main loop. Whi*n tbe ^entejktift[ parst^r hai 
cQcnpLEtcd^ thp Ejntaclic dcFELiilt acLioof arc exEcntEd, followed by thE EynLactic 
we||-formednfrisiiheffcft Aftei the fiyntartn: prc?tfsflirigi the gemfLfltK rfetault ^ctloni 
and WHill-forinwInefli <?lKtk^ are called. Onft thi* pohit is ri^athwl, tlie tnlli* paHt 
i^ finiEhEd^ 

«3 



E3 



0^ phTABal bctlAU- 5E:LECT*^ HQUH 



1 ; cQt^^r;^^ CiL^E 

chtldrfiD: <J: mDrph&JDBi HHjIAEUJ 



1; Cttflgc^ry. CASE 

cbildran: G: mn rpbema [ KAEIOI RRP Jl 



1: aDrph«me^ «ABS* 
category: CISE 



as 



pn>jectitin7 : tflL 

ctiildrfin^ prajACtion?. NIL 

oorphaiDfr: YIEIRIKJI 
CatQ^ory: HDUll 

prQ] action?" MIL 

nKhrph4ii4: yiFuunu 

pmjaction?: NIL 

iDDiphong: KJ.nDIElRPA 



pro j fl c t loa' L T 



es 



Parsing the Precedence Phenomena ' 

In tKe hktnxluj^tion tn xhi^ fha[Ht?r I ll?t^ the pheiiomcni fi:Hr TiVhkJb \hr paitcr Li 
respQEifiblo^ eeg Lable ^^1. UncLcr IIie] h^^in|; for precedence pliGnDmen^ tEi^re were 
three area±^ com pos i Lboo (far Qoun^ v^rbd, Fuid auxiliaries), con ti n un^LR rjiji^ phr^H^'^, 
nni] auxiliary pLt&LthuciiD^. fa this K^lion I [Iucje^ bcw Lhe PS paxsEr tiandles thsaa 
plienomenaH 

or thE LhrEE f^lE^n^ dF Word compceitloii^ Ihe AUj(j]]Br> hu alrEad}' bcon 
dlscLisa^l Nominal a.iid i^epbal roinpoRiLLOTi ari^ quite ^Lmllar, 50 only an ejLimiplcoT 
the latlex will be given. Consider Lhe VErb in {H}^ 

(B) <PiniT* MV 

As w]tb \i\^ K-Tixiliiiry e^ikcnpLE ^bovc, wc fitari Ihc parbcr al lhe lejcical Le^El. In 
kla niftlr loop ItHrallf Ibfl morpholopjlt^ftl p-^r^.r to t>t'i>c«3* ihe verb stem, j^bpifAr Thsa 

PS ai]d SS for punia were ^ven in fLf^urE \i 7. As no ^ctionB ina> firE, Lh? lexical 
[o&p tlH^ritou. ajid tlic tciLtr- tltmtnt^ mtp I& pcn:j>l. lift cocttpulcvl lEKica] entry k 
^vcn in Jig^re ^Jb^ and the rEEiilfln|; Etmcturea are ^ven in AgurE ^.17^ Note th^t 
ihia elEniEnt doEh nol pTuji.M.L into SS on iU own^ 

(pe (actions Clerical linjact varbH)) 
C^ji <dai:4 (tbnsE] n^npa^t J ) 

[actiQiis 4lIcfinH& (cd9e-asgi£L abac^lutivg))))) 



Figure 3.16. 1'be coEnpuied enliy for 



0[ lezLcal actinns: IBJECT: VERB 
tDorph4m»; AVI 
oatngopy: TEfJSE 



B5- 



Fisurea.[Tr PS and SS for mi. 

Wben iKe morpboLogira] PS is entered into lbe]e.ucal PS. tbe inject operatian 
of tliE tcD^E oli'ELcnL JliTJi In P£l Mhi& ca.iiBes the VGrb Btc-m and the tEn^e cleinent 
to bccoirn? sibling uiidcr a bIclj^e [lEe. In }\^^ the Byi^ta^tic infciT nation of rni ii 
ftddtd to tliit of pi^fi^ii- Onw l]ie lidflCtfa ^tion jMtia Ihc otbcr n^tlonB of lhe V 
node^ it acts in coecctI w\l\k tiiE rrmainin^ complEnent acLioD so as to plane tbe 
casfl-aasign action foi tbe ab^olutive ras? In thi> ?ero-|eveJ projf^tlon of t fan* verb, 
Tbe TEtiuItiug strLlctiutQ ars sbown ed ligcu^ ShI^h 



D. categ^rj. VERS 

chlldriin- 0- data: CflHJUClTlQB. 2 
niorph^[D&: PUWTA 
C^tfigorji VEflE 



c^tA^ciTy ■ VERB 

TH£TA-iSS7Cfii ICEITT 
cil€gi>rT. VEJtB 

TUETi-LIEfKr PATH 
THETA'ASaifllf- THEHE 
data: TENSE: ITDNPAST 

7HETJi-nr}LE2^ UCEEiT THEXE PilH) 
UENT: EEl4iiTITE 

THEME: (esdLurrvE 

PiTH. D4TI7E 

SUBJECT^ AflEXT 

OBJECT: PATS 
nciTph^n-: PUNTjI 
C*tB£4ry: v^& 

FiftUn: 3,IS; TIkt PS tind SS for punla^mt 



ThE] Bcccnd jLeiu dfi tholiat oTphenomf^n^r-HiArffrii^ th^ pTi^n^^J n£ of COAt] ti Uouh 
H^^f^phrwfS- 1'hi^wM<lt:mr?tiStr*ttd il] thH_-dlJi:[LSlionof IhGibsoliiLivecasfl-Tnajlicr, 
eJjovc. So, LIlo r^^ccLiininf^ plbrnoitifiion l<r W piPJ^nSed 1* ftM:^iliiry iHWitiOiiiuR. 

'Thv- rkUKlli^rv ^inpf^re iri ii^^VH*riJ HllffrciiL fcrtnSr ft m^ty h^ a clitks ;i ^tcird jdIh 
itf^lff cr in a pbra£G bv U^DEr k iLL^y Ur in ^hilifr tli« lirnL gt M=:rQnd pr^lMoiln Ht|| 
hol^J] iiJiLiL[f4>i^laLE[3Ei5 oniL^^uxiCiary m^y appear in ^1 of tbego po&LlboiLS. ThGrefcrc 
a d€:inoiL^1 ra lion of tbe pj-rspr'i^ nbjlfcty to liirirlle l\\i^ [th^riH^tUlH^nnn vntuld rei^Ulti'lti 
g>jL^ti^li^G iKt. Hen? [ pr«ient eLti LDclEcative ExainplE; cnQro l^sis can be fbnnd in 
tiae AppetldiXr ConelcLfT (9) whl^h Ie ^ miiior v-4ifi7itio]L ul ^hi: ^funplnz fcntcncc^ Ln 
which the auxj] iaT> ba^e^ ^^ h^ bs&n rcmovcNJ 'rjieeqiuvabnt input r^pr^eJIt^tlon 

1-ElLG-ls-^ Ubo-NONPAET child-EAT boomeraii^ 
'T will take ihe boorfieriing itti]i] thotlLllcL-' 

(lU) (((EiaAJULU RLU RECJ. RLA) J ((PUSTJl RBI)) ((KimHU KU) | ((HAFLLrl)J 

BecauEC the au^liary word bc^Liu h^^Uh a nomiinFJ agrcwni^nt dltir, it inu^t he 
enclitic totbe preceding wcrri^arrl lli<-r^r>r^ m Wfj>n-I [kWllioilr The PS f&C thfs len- 
t4i]cu ]fc^ giTci] in fi^re ^,l&. The hey Id t h^ Jii m plicity cf ibe weU-farmedneaa check 
iw Xii^ action aujiliary-adjiteni. 7'Ulft ^tlon, atld tllO J<^^lfftl wrJhfontl wl itPW 
clKtk th^l Llhc accion luui fiTcd^ ^uarai]tc« cJkLdF^Lion whei] necessary, ttius ruling 
oLjifonwn auxlJjiylJ^i^lFomHispj^^tliiglti lirur isoslurwi. t'liflKoMrtoti?, filfLC^?^1l]tilia^>' 
E^jz^CHriity docfci ]LoC coiFibLnc lim audliaxj witb oih^r trees in PS [^i €iilher the lexi- 
cal, phrasal, or uenLential lovek), llifl tj^^k fw llie SKOnrl jJWltLCpn Simply C0il5l*tfl 

3*^.3 The Syntflrtk Parser 

As mentiGiied abc^i^e, Lhe Ryntajztic parf:?.r |a nr^lcfl wLien^v^r a kjjhit of PS he^^ffl^ 
u^nUicLk^ly relcv^nl. The PitsI part of this section describee ibis mechaTiism frcin 
the point of view i>f thes^ntacilc parser. Att^^rvfrrds, iKe mair pacing algorithm La 
^v^n. Fliiallv^ the algorithih \s ricinOElritrENtHfd on the y^htaclic phenonkena bandied 
by tli£ parser. 

Synt^clit RulcvHiicQ 

Tliere are thrtt waya Icl wIlkIl a unit ^i PS miy bctome ByntflcLkally relevant: 
prtjjMiLlun ijiLoSSt si^l^^'tiorir tiid liLJectjnjn. Tlitmain uiL^hftiLlJtti ti. proJLHrtlcm luto 
syntftv, EHerforncfc! by th^^ Atlion pruje-^t-intij-33. This rauline fret prajectB the 
item accordii]^ to its Lexical LiifciTiiiatiaFi and tboTi adds the troe to phS- 

The K^'cond ]]3echaiLLEni hi selection^ which is ustid both by aoxiilajy acid non- 
a.iixi]jary eleinei][s. This acLEon is EisfKr v^hen Lhe orderiD^ of orie rEempnt wlL)t 
reipert tL> another haa a ^ynLanrrK elfrrl. ti pec lineal li". ih(^ selft-tor w la^n^n tn be 
t^^ predlcator and the 3electJ?d is taken to b? one of ila ar^inenLi. For exainpEet 



D' phraEBl actioEBT SELECT*^ BDlTir 

<]1iildr«i|- 0- lECijpbeiDs: EfGAJULU 
c^Ugc^rj: HOUN 

1: mnrph^mp: B!-U 
citagory: CJi5E 

l2 «t*£Ory: lUriUiRY-SUBJECT 
clixIdTfin: Or data. SYLLABLES: 1 

catafiory: AUI I LI A ftV -SUB JECT 

1: data. SYLLABLES' I 
■orphfioia . RLA 
catflpiry, *ui:liiei¥-datlije 

iOQrph&iDa[ POBTi 
cttBgory: VERB 

1; BorphemA; HKE 
cat&^ryj TEKSE 

3^ ptiTBsal actiariB: SELECT*: HDUM 
cat€4ory: CASE 

catfigary: EfD'JW 

l^ j&arphama' RIT 

4: phrasal RCtUiw^ SELECT*^ Jmun 
^ata^ory. CASE 
4ihlldr«n' 0' marphtme' KARLI 

Ij li»rph&iii»i 'iSa* 
catagQrj CASE 






cajL'-inarkors SL'Lcrt nutinE. Tikis a4~tLDn c^us«£ Lbo^yntacrLc CQUCLtcrpart cf Ihencmn 
iA> ho an argunkFnt of lIl? cnrrcgpoiiilkng ^j[}\^ ])lkr^R& Iq HLTuduraL lermB, llii^ □guti 

InjccUon k0 lltelEiat rlh^iiiLijI lr<Mn PK it^ fJS. With tlilft ai^tioti ilse ^yittftrtlc lifw- 
niAtl^n cr ths injector ja Added La UltlL of ~Lhe injected, An exampLe of^hia CDncemj 
\h^ p4tr*iiLg ufliLilcclcd vti^>Sr WIlcil tti-^ tciiit t-leflitJit cliticiM^ to tbc verh sti^niT 
its i^ynt^tLc tnfoTmation je added La LHal cf tl^c verb'a. The lic^nain^ action fmr Ihe 
tasft-rmtrklns ol fthgoliitlv(- fase ia tli^ n^s-iJ* bU of (ji form At ion that is added The 
B>ntaciii: pani*ix will tlitiil bc ^bW tft add Ihv cas-C-ccLirUtiS i^^tton lo verbal JJroJw;- 
tjon {lti the 7ej^>leve] node}, beEiauEe both ibe conpleAont adicm and Lbe Ixcanae 
attion for tbt c^^^^:- mAJ-lti Qg irtion i^te ]bj |Jliu-u. 

After any of tb(>M>4iJ:l]cinji hav Ext-4.uLr}d, IIil fi>nt^LrtlE; par«tr Is 4^£^«h tO«(^ hft^tiy 
furtll^^yDt^tlc a^Llon^ c\z\ apply, aii ii rn^^uh af th^^ uddirinn of Hyi^tartic "itruntikre 
DT informatioci. Tim iii?>Lt scctirm d4^scnb4}ti tbfl workhunt ->f tli<- syntattlj; parnn^r 

The Baeic Engine 

Tll*i parflinft fd6i>nLhinT gl^i^n in Csble 3-fl, loopn thTn^iigK tlie ijneVect3li>d ictioTifl 
of tbe Eynlacdc ncdec to Ee« if any inay be exccuL^. (LiLe Ibe fS engines Lhit 
parwr ^so employe a qet of vinfiatl&lled pirerilrateK for plPfienny'fl wtlte.] Note that 
preinouEly eKC9 Lii] g actions axe alsocheckjz^ b^caufc ibe aewly added sliuclnje msLy 
provide wgumoiits fl.p|iropTiate for th«u. 



1- T.oop [n'Pi every nnexef qtpd Firtqon in S^: 

if Lbc utlod applic^ii cittu^ to di «ibll[l£ [>t io U tr«« 
in ibe strurlur^^ ^PP^y tlif action. 

3r IT any w^tlfflift hflve flred, try the above looji ig^n 

Table ^fl: Tbe Rynt^cLic parsing enj^ne. 

There arp four ^yntac Lie ActionH a.va[Lable^ ae luted in table i.Q. The fiTyl acLboin^ 
caSB-4BBi^H pH^rfornis tlie second biJf of fat* abSLgnnicnt. Tht fiwt *lep, fasrh 
niaTktTigi jieed noL be ^it u'Lion^ Bmce tbaL fpaLare la Inberent in all of the ease- 
marker^. Th^E^t is, tl]4> SS iJ;iLii£i], cABa-Hark^d, L^ slr^ri][l in tli^ kxical r^itry U^ 
earh nf t|i^ fju^rnarkfT^ (fnr their own (^Bfie, ol ro^melL The ne^t Livo FLctionF^ are 
ror^-aEEignmenL Tb^ second such actichCLn theta^lin^^ ih zu^^^ crnLy tn^t tlic ^Ih 
tf-rol*, A4 it la ftsglgned dlrecily by the datTve nrAfle-niarkej"- The Iaal acHon i^ for 
-tfHEfal ar^ment L-aking. CnrninLly^ eI i£ lucd only for the aiLxiliary ^hjcb La^Les the 
^erhft] prrtfOH^Llon aa Its -^cpIp arftumenL. 

We now return to tbe second step of thp parsing algoTithm. If Any acLioQS are 
executed m the Eoi]r£e qE tT^ Versing; th€aeL ol unE^tisfied piedicaLes^ th4^1oop Ls Lried 
DnC€ 4£AJi3- In thJE way, a11 the a^^Liona thai are Jipplii'able due To tbe jntrcdu^Lion 
n| ni^y^ isyiitftctic t^triwritre T^'ill he eKfli^uttd. Tliis lo<Hp al-si? eJljnlnates i^ny ordertng 
dfect^ of the actions in that if one a^iLtop ta depi^ndent on anoLber^ it doeBis'L matter 



thflta-LiDt (Tolal 

raati^Le Lhe urdtrLn^ GfliiLl of twoaftlrfcilft tti^t m.Lj- befl.pplicH.bfe tDtb€Ba]iiG codpitt 
tbeE-amn? Iimo. Tlkc ]?^rS4>r A^Lini^9 thi^t At moflt one euc b aclion Will bc^pplI^Ahl^ 

As T^'itb tbf P^ parser, tbere are also rouline* for [fcttfofCtlLllft detauH attioiiB and 
rKg rfciT i|F ib^ Wel]-fonn£dtl4^se of l\lt ^yntaicLic ^lucturc. Botb Dfl^hnsG routine ^UY 
Uflcd by tli« FS paj^r ChDrc it has ccmplctcd t]ie H^ti^HtLal level oE pracESBing. 

Tlir iiJily chrauLt Mticm oF lKp ^ynlactk parber ib to flupply ad AijXJIlMy wrtpd if 
cue IH m^L pT^Bertt in t|ie input Ebvini^. Sucb an ^nXJ]i?.ry con^hts ofth^ null ba^e aud 
null agrwrnj^ill rlitJc:^, wbitb COJUain t-be default i nformalaciEi 3S mpjitionid ibovt- 
Nato th^t pl^]jl£ the handler lor Lhe ^oiv Auxiliary in tvEiliLX ^miil^ti^ the n^ 
for Llip parser lo j^cBt where Ibe *iixibftiy i3 to ht^ plai^n^ m llie Input Bentence^ ]t£ 
pLacen]i?ELl dactj iirit m^-Uttts ^>tdv ItB ayntactit (.^rd then Bemintic} GlTects couMra 
the pareej"- 

The well-formediies£ tbeck Iof SS i:oiLSl4H of nmkJn^ Bure thitt the atmctnre 
tontwtis eXMtly i>ne ti&t". ThiB tbeck auhBomeB tbe Cd^Ji FiH^.^ In thfl-t rn^nns will 
]iot get Jinked Lnlo the verbal proji^ctLOii tliili^S ti\i^y ^n^ ai^ptoprtately marked for 
T-a^e- Noi][i€ that are not inajlicii for ca&e wiLl nzmam af ^tpir^te tree^ ]■? tbe 
Btnictnre, ind will tbuh l>^ lUggi^ at Itl fiTFOr hj Lh|g joutiqe. 

Ponjtig lh« 9ynlactk Pheii«m«iiti 

Tbc parser C^iaa htuidh> tw^ lypc* tif s^nta^tir phenompna- deternunation nrgratiL- 
m^tKa] F^inction and free piiFase order. In thit Mection I dJS^USS how tbe ByntacLiE: 
parser fpcH about tbit. 

Thci majority of the work In deiermlnLnj grajninatical function is littMrlly per- 
formed Iji tbeevecutlon of thp lexical rules, at described aljOVt. The £yntiu:tlr paraer 
jdfit makes thom slack by allowing adaorE^ to op^Tfl-te only on slblinfjs, ^knd not Od 
other part? c?E tbe Stricture, b thie way^ the actton^ plact^d in tbe 4t tu retire ^t their 
appropriate kvclfl wlIL in fMt, cause t^^;lr arpimenLa to be plajied In th(> pioper 
places. Suhjettt wjtl bt adjwned a^ slbllogB of the first-lev*:! i^rojcctior, Mid ohjCMita 
will be ad|c^ned next to tbe ^ero-kvel prdjerllon 

Considr-T the verb Btf-m prmta once aj^&in. Given iNs ^-grul^ Anfi tbe lexlEiaE rules 
ibit^ver th& p^-Td^f dftertiJces tbe mapping from case And S-r^Id to ^Tammatical 
InnctiouH hlh shown in tabk3.10r This mappirig ia implemented bv pLadtig ^he cajE- 



'Tht^ n^i'tlMNf ^^uni^ un^iiitiLxiioHU L7aCjinLC Ftmclani Of ruiirs«, t]icr£ ui itmrLiiTd \TW\h\- 

E't^y LEI W^pui — UEB anFGtlicr nitnnl L^lipUl^. Thlh ^ll^TtcornLHf dF the puoet 3b dliCU*H^ in 

89 



ani3 fl-.iHilioiiii at their ri?s plh: tivc [j^^clf-icL Itic ^fti)wtL&tir FlpjtP 3.18, ibt^SS f^if t]ie 



CH2G S-Tc\n ^r^inmBlk^ f^siittku. 



ersatLve 


a^nt 


EubJDcL 


absoLutiVEi 


Uhcni^.^ 


e?bjdcl 


dative 


pEtth 


ohjpct. 



T^hle 3 10: The czbe/BJGF lUBpping Eor pHi^tft, 

The k?y U:} pRiC^i^Lng thi^ Bccmnd pheocniciioii^ frea phnue ord^r, Dcf Ln bhi? lu^k 
of precedeDce inronnaticm in SS. W\i^\^ the j^yntArtlr e]h|v;iiie BMrchea Tot polenijBl 
AfgumcntST ft Irfl-VErfififi the entire BPt ol Ircca^ reS?^rdi«a oftfn? ortkr in wM^h ch^'y 
wpre addi}d The ttHnrh ilfti^ Involves prevlrpyaty idioined treeE, t.e.^ tiblinfis of the 
nodi? iTonL^lT^Ing the artion to be c^ecuLe^. SkbLiiigti miLflt ht cliecked b«au9e more 
tb^Ji oiie actLoa may apply to ^ ^Ln^t il^4. 

Consider A. permaTatkoi] of thEEampIr" fig]>lmCG^£hownin4ll). Tn order io appre- 
ciate the feyotiCtic pajftftr^ lirilfF^rpiiiTe lo precEdence in the inpul htrii]^, Vr^e (wmi 
i>tl chr PTlLry cpf the inJLecled verb intc SS. When it prti^XlH Itlt^ SS hU argili^^nt- 
lakini^ actions (e.g.f cafl*-tfljignj bwnme manifest in i\E piojection. The main 
toap cf Ih^ cyiLlajiilc] {^n|^ne then &tarU up^ soarcbin^ Ibr rirguc^iit^. II Anrij ^ of 
the arfiiiments that ire t^resCnt iHd fjin*^ thean to the verhaJ p roJpctJQii^ tbe oe yet 
IMientored actloruS t^J^ not joined, and the acLiDiis perform]!];^ iLoir adjuHctiOi] arc 
■inpLy left unejL^uted. 

(Ill tt{lULFLLr)> «Ki MA RLA)) ((PlfflTA RJJ] » ((KSliJULU BUJ]) 

I b 

[ft (H) thn^ ^prb pfiters after one of its ar^mentfl, t^Qrli, Fiit altt^v pfOJ^tJ^d 
Into SS. Therjjforei only thp arLloj]j> for the object may ftrc. The inlcrmwlii-le SS 
LB shown in ligurc 3.-20i the »yilt*ttk Stcnftiire of the auxillaTy Eias been omitted 
(or brevEty. When the TonTth word, n^_^-fTJ-r'u, cillery, thi? HUbji?<^l nCtldlL^ fll^i ADd 
when th^ 1^1 wotd projects tnto &&, the iDdirent object actiom lixe^ cDcnpLetiDg the 
parse. 



3.2h3 Semaniif^ Tnt^rpretataon 

Ab mentioned abovef semacitic ptOtCHP-Lrtg 18 prrf&rmed on Sil- After tbe syntacLic 
ftrkjctnrE bas is^en rbe^^Led for well-romiGd Df^^ it isi fb^rtkHM] for ^i>ni4ntlc \wll^ 
rorm^Ltl^fn Thf^e rLfvLs itre ^lescribed betuw, but Jim 1 be^n ^itii tlhe ]na]ur 
semaiit]c opeFatioUt argument idf^ntifhcatLun. 

on 



proj&ctLon'r NIL 
datiL- l-RciUHENT- VERB 
catajDry: VERB 
childr*rL- projactiflB?; T 

TUETA-JiSSTCffr A{]EWT 

citafiory: VERB 

chlldrea^ projectioD?: HIL 

CJiSE-ASSIGWZD; iBSDLUTTVE 
category' CASE 
Ehlldxen: projection?: II L 
piarp^uDB i KJJtLI 

projection?; T 

data: CiSE-HiRKFD^ ABSDIUTIVE 

BQrph«DLft! *ABS* 

[;st46<3ry: CASE 

prfljwctiih?. T 

actioEiBE TETETA-LrKK: PATTT 

data; TEta^E: HOnFA&T 

TEETIi-HQLES: (ACEITr THEHE PATH} 

KtVT, EEl^ATl^£ 

TtTEHE- ABSOLUrrVE 

PATH: l3*TItf£ 

SUBJECT. JiGEirr 

□ BJEcr PJiTH 
EDDiphsmfi! FIWTA 
categorj. VEJIE 

Figurca.SQ: Tb4> Eiiterin«diiit4 55 leaf (11)- 



91 



ArgHiTkCrLt Klentllifat^Qn is ihe procesE of a££oc La-Li np^ nouitK ^'itVi tliO Si^m^NLltk J~ole& 
they TlU. TIl^ olj^ontliiEl fOJ afgtjmi:nt l<leiltHli:a1ioJi is fairly simple- Fur eacti rola 
La CJtfr v^h'^ ^-grid, tb4^ vf^ibal pnajecLkui eb &€ajr.b^ fcr ^b^ ni]iiii^ lliriL Il^Vi^ Imt3 
Bdaieoed that role. Fhjt Hi\;wlipliS Ihn^ Ln-uppJnfi oT ff-role*, to houilh for Lhe tajnple 



PITHt projecti.on7: PIL 
UiO^ptLeDLa: KITRDU 

TtfEHt- proJ^C^lc^Fi'^ MIL 

AQEfJTi project iunr^ NIL 

data- PERSC5S: L 

mfhrfhhamfl^ NQLJULU 
category; ta&UN 



Two -^emaiLlir we]l-forinodneE3 chects 4t^ perfortntd hy iht PH^h^Ft fAfli HftaUng 
with a difierent pari of Iht duJcihaiy. Ttui lirsl J^JiSTLlfft that th^ auxiliary biM ie 
^pproprl^t^ for th^ tene^ of the LnFected verb. Tlut check \e r^Lhcr &li^£titf<:ir^IU~d: 
theMEitenre is well-fonned if lhe \^^^ of thtt&nse i^knwrt i=: a member of t be Mt of 
the allowable Lciiaui ol tln^ bssn, hor Ktampkn the Hajnple senLj^ime pa£EJ?5 this teat 
b^aU0O Lt9 tejai^^ non-pafiL, js allowed by Lhe hase^ i^ If tttn^ thei>ther Imp^fectlve 
h^^i '^ju, were iLsed it would be il]-fon3icd. 

The second wdl-fciifjnKlil^ij th^k f^oncems the a^reemeDL cUlks^ atld COil^ifltn 
Itsdf oi two p^ris- The RibL part checks licensiiiE o\ fliti^S 'I'liat In, agieement 
cLilacB Etre graLninlLlt^ Oftly If there Axe corr^aponcL] ng aT^anientji In th^ s^tit^tM'^ 
» ItcenEed b> the verb. Connadtr Ihc vj*jb sI^^JHh jjun^o- It selecLa thiee fl-rolcfi 
wlLich *ri^ inanEieat aj subject, object^ and luditett objett; all thiv^ arpimflitFi are 
IjcenFt'H. B,v th^ rnlf-fi H^ojifPinlnRauKiliary TesittratioiLn Ih^^ubjcf t Wid the in^llTf^t 
object (f.c, the Cf^^tlvo^nrL rULiv^ rAfie j^hra^eB, respectively] mnst EiO r^gtston^ In 
tiieiLuxihaiy with cliLics (whii-h ]isa>' QOt hepfiflnolftgltajiv ovpit, howe^r}. On lhe 
Utlii^r ^^ari^ consider Lhe vi^rb slem^ nirina \o be^ to «it'. It i:<HfCt$ a single ^-role 
which ehow^ np a^ tho ^hibji'^t. Ae a tesuIc^ the auxiliary may not Appear witb iin 
L>bjwt diLLC or a dative re^^^LratioLi nibrker. 



TEl^ nlher p^t of tbe Agr^emEnt check matches peiBon ^ni nuiril^r m^orm:it|on 
nf LhecliUcB and ibeflrgunieDts- Tbla uln^irk tE algt^alnipk" tllcpertKHii And ilumiHun^ 
TDUst agree- '^ 'I'hwe IS a slight twist mtli ftc>nilii-Jft with no o^arL numbflr- mirlc rr 
tln^y malch eiiher Binflularor plural nuinbcT- 

TEl^tg ui another twist in lh]& prccessin^^ naineLy, handling null clillci^ And null 
acaphora. Null chtlrfr liavp ilefeult viJutft irf l|i i rJ - pf rst>ll Wkd slJlftUlif- hi tllni Ci'OcLt 
of » null clitic, tli«^ VSrlUHS £^/« WriQVed knd matched as ofual. Null a]haphQra 
ar£ handled diEerently. Given the ]ptTaE«utentj^ pTocej^smg n^f tfi^ parser, th«e 
EleinenLa are confilderfM] a_R w]1fl-i^an]_=: Jot lite |Ji^rpi>M^ <ir fifttJ^tnen 1, Atiy cLltic will 
r[Uirt4.l3 tl]cm. In it paner tliaL handleE mor^ ihau one sentence, the clliicB will h^ve 
to match Lhelr reli^rentB, juRt aji lit the casp of ov^rt aTgumentA- 

DcmoHBtFatin^ ttie paiBcr^B perfDrmaiice on Lh^e ph^nom^na 1h bi^t rion^ i^Uh 
an exh^U&llvt t«St 4>f tlL« iK^elbllltki. SucIl ^lll:t r^n hi? [iMind an tb? appendix. 

3^3 Parsmg the San^ple Sentence 

Below la A pitr^c tmcc o( tlhc sample &gu tence^ (I). Each line corre^ponde Lo the 
execution of a-sSiigle fttcion. Line- begin with the naino o[ tl^r [i&Jc [Herforcnin^ the 
nctir^n. Jollowed by the plbontjoglcal Ipv?| M which tJi? tcHon lOOk placfl (lor PS 
^tlon? only}r Lim^ c]hd ^^Ith th^^ action that e.fecuted and its ai^aments. 

Tlie (iHt livt ^irtlEjtUj [Hitstd tlni fttst -Wordj ngojuJu-rfu. .Vpcyu^iA projected into 
SSs <tnd the BynLactk paTser then projected it Accordinf to ILb lexical ijilorrfiatlof^ ^ 
^lUch iv:^ ft}t ZL'VO kvela. ff^u v^Ehs thn^n pE^r^ed similarly. Once both morphemefl 
Mtfired PSi thej wers corubjnecl with the e«l«Ct ftCtJoji of p'^_ 

■C*JULU (HDRTHDLOCICAL): PROJECT- 1 HTD-SSQ 

HGiJULU- t^ftOJEicriCUfS'O) 

PLlf lHanPEf{]LQGrCAU: PnDJECT-lKTa-S5 

tlUJt PFLDJECnONSd) 

tUl (LEILICjIU: SELECr(Efnmi) 

Tlhc n?xt set olfactions paj8«d tho aa>:jLary^ kfi-rna-rriQ. Firet, i'G projected in Id 
SSt projected tiAu levels, Jiiirt plftTod Its srgym^ht sr^tlor (which latfr e>rw:utH^ or 
Mtt: vcrhdl proJEiciaon] in its Bpecifn^T petition. Ihen the: subject clitic, ma^ entered 
and also projected intoSS- Onf"? 4'^rj pnt^rfd P^, ^ wa.^ ahle to first n^n^iblne witll 
|L and th*-n iiClect it, 'wWh right- adjacent 4ind aujlliary-aalact, reapccindy. 
The cLf^tJv^ rtgivttMion clitk^ r/a, we£ then parsed Bimilarly Lo thE] fuhject clitic. 

Ri (HDRPHOLDGTCJIL); FBDJECT-IJITD-SSr) 

KA; PElDJECTItiMStZ) 

KA'r SPECrFIER<(AFL[}Lr}tE!fT . VEflBJ) 

ftMA (HQWHOLflGICrit-J: PHIJECT' I EfTD -SS O 



T'liA fririhiu3FLt]0H hti nul qiiiU: rtf^liL F^ir ^ ^ull acuDjal of Warl|nTi >p;n.TiTtrHt| InL it iltilt f^ir 

Lb?ifiniplr ciflGC dF tlic papMT. J^w JNa^^i] txf 4 intji* Tit^rly f gpitpH^ ih^-riiRiwi of lUl* iBiHCMr 



FLNiX F^MJE^:TIaN3£01 

ftNA (L^KICliU- J-UIILIAHY-XDJACENTCJ 

Ki (LEUCAL): RHjHI-ADJACEBT( (ITJIILrARY-SUBJKT AUULI JlHT-Q BJECT 

lLtltJtrAftY-&ATTV^)> 
KJl (LEllQAhi'. A[li;rLTJlFLr-£ELECTCl,UXILrAHY-SUBJECr> 
BU irtORPHQ LOGICAL): P FlQ JKT-I JTTtl- E5 1 J 
ULA: PfH]JECTI0Il5(0) 
HLA (LEilCALJ- AUirLIARY-ADJACEHTC) 

HCCJl <LEirCAL)[ HeHT-linJAOEWrtCAUIILrARV-nBJECT AUXILTARY-DATIVEH 
KA £LEiriCAL>: AI;);T LI ABl" ■ SELECT ( AUH LliRY- CUTI VE) 

Tho following set nf acLioDE poxsGd IIie] vtrlt^ pwitfo-mir First th^ ^wb fltem, 
fn^nJfl. projecled intc SSr Once t|ie verbal projectio]i enlcrwj SS. Un: ^iiXJI]tLry 
(bn^ed by trt) wai able ta attach it in its ^pttlfior pcsLtlofLT by th*^ arguaaQt actinii- 
The ae_lL ^ictiuild H^ tfi^ "^PFb placjn^ it& ca£e- iiad ff-acliojui id Itfl stnii"lwrH" N^l^ 
thi^t tlie fttaigrment d( er^ati^e ra*ii: and tht AgWIt ?-rc?le Kif^^ut^ ti41w hprausa Lhe 
erj^atlvcLy mirki^ roiin? pgrjjuhi - rJu, in piobent in SS. Thtj utlitT case- atid fl-atClociJ 
of ill? verb muat wait untLL thear^ujJiH^ats ap]n?&r Jn SS "I'he la_Rt acticrn for parsing 
Ih^ verb ccmcernt Iht L^nsc clea^tiltT HM", which conihiiics with lJlg varb sLEinh and 
iDJectE lU byi]Ul^ti4: jiirorinii-Linn IdLo the verbal prnjectjcn. T^be^ulilod iafon]l^tlon 
ofint^r^fii L9 the acticn Lkoniiang cf ahsuluUvc: ca^G AAi^lgnm^nt.J 

FUffTA (HDRPHDLOGICAL)- PROJECT- 1 EfTO-BSO 

PUHTA: PEiCIJECTIOSS[^> 

FU[n-|l: 3PECIF[ERi(TEIETA-i33rGS . J-GEETDf 

KA] iRCUfEBTt VERDI 

pUEn"il" SPECIPIER((ClSE-AESI<jfl . ERGATIVEM 

PUHTA: CJi3E-AS5It;2fCfJtal,TIVE) 

PUMTA: THETA-liSSICErUCEBT) 

FUHTAl [TDnPI-EKEiaT([THETJ,-AS£rGH . THEHE)) 

PUECTA. CDnPLEPtEiaT((™^li-LTNK PI.TM)) 

fUHTA. CDKPLErtENri(CASE-lSSICH . IJBSDLUTIVE)) 

RMI (LEIICI-L); IB JECTi ITERB ) 

TKe Iburth TTOfd, tnrJu-tir, i^cw |3are«l wjth tlie oz^ecuied actiona belry^_ Tbe 
parre hert pn>JW<l^ <|Ult»^ lll^ ihat rsf the &rst word^ n^ojuJu-rfu. Note that Olice 
tlie daLive caae pbrdtp Hrciti^nsd SS it ^HS llnlfwl to Ljie vi^rbal prDJeckioa via the 
theti-llnk acttor of l|ie i^erb- 

IfURDU (HDRi^hCLd^^lCALV PFLDJECT-rSTQ-SSO 

ECURDUr PRDJECTIDNStCJ) 

KU (mRPHDLOillCiLl: PROJECT-IMTO-SSCf 

ECU. PFLOJECTIOItSCl) 

PUtJTA' THETA-LINE<PATT5> 

KU a^KICAL;. SELECTCHQUtf) " 

94 



nokiD, ittfrri, WM ioL«led by II^h^ f i=e-Tnarl:eT. 
KUILI ^HC'TlPHOLOSICiL) : pilDJECT-IMTD-SSf) 

•ABSt iPHlUSflU: PflOJECT-IhtQ-SSO 
•AES-i PJlOJEGTlON^iLl 
FOBTJll C*5E-ASSIGlfC*BSDLirnVE> 
PUKTli THETA-flS5rGfJ(T>IEWEl 
UBS- (PnHASiLj. ZELETHmaVi 

of ihe verb's ff-cftlcB Ld wonU, *Jl^ ff '■eii in fnii'™^.^'' 

Tlip y recede rP Rlnitture For ser IpdCB ( 1): 

□ ; phraftiJ. acvJ-ona: SELECT*; HDTllI 

cstogorV- CiSE 

chilican: 0' aorpliuno- fJGAJULU 

citBgnr?: JUJUN 

t L moipHina: BLU 
CJLtaflorj: CA5E 

^hll^.n. 0. Wcl-U^ns: idXILIAElt-^ELECr; miLim-DBJEn 
da": SYLLABLES? 1 

CAtfiaorj; aUULIARY-BASE 
1- dati: STLLI-ULES: L 

catB^ciry: jlUlILlART-SUBJEn 
3- data^ SYLL1BLE3: 1 

cJitfrgory; iUirLIAfl"^ -DATIVE 

7" catflfiorj: VERB 

childt-nJ Or 4ata: CPNJL^GATIUN ■ 2 



catena ry^ VUIQ 

1: dAto; CQEfJUGIiTIDHj ^ 

catagorv: TEB5E 

3: phTiflBl atl^tOho: 3ELE£:r*j aOLTB 
catefiOTy: CASE 

C^l«e4>ry . EfDUM 

categoTy- CASE 

4: prtr^sal a<:ti<]n£: 5ELECI*F NOUEf 
category: CASE 
children: l>; QLCM-pbemfl ^ KJlFlLT 

It [ni^rphBBQr *iBS4 
catdgi^ry^ CASE 



The &yiiLft£t|i: iti-utlure for Bcatence flj: 



projaction^i STL 
ctttBfiory: iUlILrAftV-EflSE 
children: priJ*CtiJ3i]?T HTL 

data; J-RCnKENT^ VEflE 

chlldrfli]: pTD]acticm?^ NtL 



dfttfl.. THETA-iSSH^NED: AaEUT 

CiSB-A5£HjMEDT ERG ATI VE 
catagcirT CASE 
cbllilrBn: p^^ J ^^ ^ ^ ^Fl"^ ■ WIL 
JA^tr PEH£OH: 1 

ECLTMEFLi (^XBCJULIift) 
[DCir|ihciiDt]_ (fCAJCJLU 
c^X^^utf, ECDUH 

proJactlutiTl T 

di'-^: CASE-HAEUtEB- ERfiflTrVE 

cBtagory: CjlSfi 

prajectiont: T 
taxflgj^rj : VERB 



Chlldrtn^ prajfictiDD?: NIL 

d^tA: THElX-iSSIGHED: THEHE 

CASE- ASS IGflEC: A&SdUJTIVE 

thildtt^m frfijflciiltn^' MIL 
ErorphQiTA . KAilLi 
ciatafi^ry: ETOUIT 

project ton?: T 

data: CASE-ttAHKEI?; AESOLUIIVE 

iDEirpheiner *AB&^ 

CAtagar;: CJiEE 

prDjfictitin"?: UIL 

dttft: THETA-E-INHED: PATE 

cat&jorj: CASE 

chlldrap^ prchJa^tiDQ?: ffIL 

cftt^gory- suim 

data: CASE- ASS rClfEQ ^ DATIVE 
CASE-MAEL&EDj DITIVE 

tmeta-assiqneh^ path 

urarphstw*' KU 
catag{?jy. CASE 

prDJaction' T 
rfatiT TEflSE; HON PAST 

TUETA-RDLES: IJlGEKT THEHE PilttlJ 

kQtM-r mCATIVE 

THEME- ABSDLUTI7E 

PATH- DATIVE 

SUBJECT: AGENT 

OBJECT: PITH 
nrorptLBne^ PUKTIi 
catagoxy: VERB 

c*Lftgory: AUK I U ART-BASE 
cfalldren t pra ] ecti dd'^ -, NT L 

norphflma: J^LA 

category; AUJILIARy-riATrVE 

ptcJKtlChti?: ETIL 
BT 



ITJtfflEnr SIfffiULAft 
mchrph^AQ. TlNI 
tat afiury t iUI ] LI flRt - SUBJECT 

data: TEMSES- ^XDNPiST) 

jlspect. imperfect 



Chapter 4 

Conclusion 



The lirst 5fctlofl of tills cliapcer dlsc^wMW M>m*^ other rurn^aii gr^nimaliral frames 
WOrka iin^ hu^ tliE]^ addii^ss to pbcnomGiRDn of ir^:^ i^iurd ord^r^ FnJlciwiDg this Lt 
tumff Lo the panor'^ e bcirtccnumga AJid haw Lhcy inigbL be i3V€:Tcomp. 

The [Hftj^Gr pr?ff€iited lierc le nol ths only member of ]t£ lainiLy^ but It tia^ only 
a. few cQUfring, Th* cIo^hsi rd^Uve w iinutlier ti<iver]i]jieLit-BLiLilJii^ ijiistd W*rlplri 
ptrrtOti Writtci] hy BinDcoiL-Lokcr. UnforluiuiLcly, the wrn^b hziB i]oL jet been pub- 
liahed, ind 1 have not ypl |i&d \ uliinre to view the nvp-tem In i>|>eratlunp ao I can 
not Mjnmttit ou it titf*. Ilftwevot. u com[paTi*on o! [h^ [numfb stLuuLd j^rovo lo ^n^ 
quite Inlercf tin^. 

ThPTej^reftUciafehaj GB-hMe<l prtK.«*ftr6 IntfLf^litflrittirrthatworkonlajigiiigci 
otb^f t-ban WarlpiriL A ^on^pari&DD of those warl^ ti bevond tho£CDpe of tbl£ thesUr 
bat I pwaent the list ol ttiH^ systctne of vhj-:b I tnow Jti tihl*> -1.1. TJtp ititcnati^d 
/Bader la jefeired to the orii^Lnal pubLicaiicnB. 



fi Abafly^ English ifa-TscrfAbuSTj 

13 DoTr'5 EnftlHh-Spftinjsh trwiftlfrtoT[noTS7l 

D Sb^p's English- Spanish tTB]LElalur[ShBti5] 

Q WeJbfirs Ft^ncli pai^r[Wr>|ia4l 



4.1 Olhnr Gramm^itira! FraiTirwnrk^ 

Thiireare many aram FtiatN'Jil fr^mi^ivork^ h<>5tdjM GovtJTimenl-BlradLcifttheary. ilaw- 
pvi^r, I shftll fanly be abit io rflsi^uSN a R^w ol thMn here. In p*rtfcfnlfHT T ^-\\i diMu-yi 
ID/LP CT^iELinar^ Lexkal-T'ij author^ Gr^i-niniaTr aad 'I^oe-Adjmain|^ Cirammar. Of 
cDjraei ih«i; rj?viows iwvbJU^E, irtJ thtrcELHre Jo aot Jo jiu^tK^toih? i^atlrefonterinr 
lh£ thi'i^k-^^ tliib dijiEUESkOTi Eocu^es only oa Ibeir abiLitv lo an^yifiz lliu ph<'Hr>[ni;iirhn 
of free wojd 4f d^r 

99 



1.1.1 ID/LP Grammnr 

II>/IjE' 6ramcnarfi[CiijCIJ ccfccitcun lwi> klnris of ruJea IL ruira (Eict^o ihe iiT^ngrdi'cJc 
[/<^r7ijrMi?rr- rtlatloiui of lh*i conEtiliacnts of tliu ^rilT] [tiar. whl|j? L^ rulefl conalraio 

EJiluaQy [^^USCntHl tn iflL^ in L rod ut Lion J undiir Llin guis^ i>E ft mricilEirf ronteKt-frcfi 
irraftiihi&r. That granuDnr f^ir^lmpl^ tra^fiiLiv^ &imt4^£cs is rEpcatal b^u ^ (1), 

Thse &r« Ihc t^o ID rul^ 4~if ihe grajnuinT. The Firjit rule Etritn?^ tiaut id S 
toJLElsl? oi^W NP antl a VP in eillin- i>th|ht: tUr hftTflibri -^tAtffi that a VP noiaaistt of 
fl. V and an NPi 41^0 Ln wUb^T ord^F- ThJa Kraitimar does [lot corLaii] aHJ" LP rulia, 

Ai mtntlonH?J In tlip inLroductioEj ju^iiini^rj of tills "surl can sulTer I'roni ira^l*?- 
r]tLitt^ i^veTagP- For examph, iIlc ID/LP jrwamjir ahovo caa noL fieiie'rate eithtr af 
Ihc JientH^Ct*cliei"n^a found in |2}. 

[J) V NP, NP^ 

Thl;^ fiflveriv of zovht^.^ se^im. kn bni at O-lrf-^ wtlli tl^j^ atrtinztuiet of the parBcr 
preealed ti4.Tn. AfU^r ftlln fmeof Itt aLniclanra reprcEtnt*! [^^l?c■c1]t[l1:15 and OAC rcprfl- 
Mtl^S hiutJcby, wliich are vcryELmHar 1u Ibt^ LP AUCi [E> J-uIcBt fWipectivply. The key 
difF€icnc«^ hoivoi^'f, JJf^ tn the 3c^pe of Ihs hierarcliicaL r^l^tiChnS- In II? /LF grfini- 
mai^j all the chlldrGiL of a slllftl^ fllbllnE mnHL be adjacent to ea^-h ot^OI, luhjwt 
Lo the liacit" piecedenre given iii Lhe LP rij[es_ Yot tin? pUfcr, OH ttie C^t-lier Tiandn 
there IS ni? iiiach reatrklioiL on tin: rhildf^n i>f ft gynLxctlf rode- ihev may appear 
Wiywliereiii Ibo Lrpu t ti ["iilE- ThUfin the l^kgu^eta allowed by the parBers gmntraar 
BitbauiitC ID/LP [i^riguagea^ and alio*' for thi: ^rrainliling isvldetli In free-wofd order 
IftJk^uageg. It ig iniporlanl Lo note t^at tlilft COVUragi? 15 notftain^l at Iheoxpeuse of 
linguistic peirpicnil.v. ll]t pWMr J9 still atle Lo represent the r«kV4nt h l<-tit^hn:Jll 
Blrutlnret in oiJer to fKOVer HPmantIr rolea- 

4.1.2 Lexical- Fll lie I icr^nl Chramiimr 

KlavanE[Kla!i'J]liajEhd*in thiL L^alral-FuncLioBal Gfajnmar|l[rea^] caHiccOHill Ibr 
free-word i^nlfr ^p|LO^K^m*^na in much the Eame "^ay ut ptt^nt=^ ll?TP- Tfie i^^y to 
imaLvxinF thcBimiLarLly hetw^jn frH4^ anri fi xed -order Bj&temh lit^ nl LFG's bllparlltr 
wpKfleplaUon ol r-d^TJ-r^urc and f-f?tT^uinrr, C-rstrHflUT* Ift ordeied by ]jrecedence 
and hiCT^nby, ^liv^i id W^^ to represent the ordercHi phciLucTKIU^ fltj^zb ^ fontlnuouf 
cft^e pliraBCft. F-s trnc tiin>i on tb*> OtJier llftnU I? J^Ot oidftn^ by p recedeoce, latiier 
IL JS vtrictLv a bl^rartliieal ^Iructare^ used \g repnrflCDt g]:^JibM13.tl^al Eitn^tions. Iiike 
tim GB j^immnr meil by thn^ parsr^ri i^^lTUcLnre iB TelaCi ■[ Lo f slruttUri' by fa.^d- 
ttiarkltigr 

lOD 



Til? L FG aikziL^ ie can b? LLLustra^.^ tura Eimplelarig^LBj^EnJ traiLshivf EenU?i]ccEL 
Tbr ^rasLmar give n hen ^ tiken iu\vs% Klid-vani' pi^pPT, i^ for Ngly^n^ba.i^ ^jtolher 
»hi^r)gin^ laihsua^L^ rrona Au^tr^la xUm ts Egijir-OisijnllArto W[hil|iLri. Tht i--E^trU4?lii» 
rul^ jj« fibown Ln {^}.^ 

TheBC rules rover cmatimtoaa rns^? phniaefiH labeled ^T!\ Sm^lc-noiir tji^c pTi niMt 
are alfo baDctliid: they are labeled 'N\ The t^ructureE ([erived ^rh c-slrHL tiiiO &r4 
ainLfptated both s^tth gfAmmatin&l lunflloii aui] iia&€, aj bIiohi^ jh (4), ^'bijuh ^ves 
tli^^Jinota-tloikfl hoi\[ for the sulrjei't. anO the obirMit- Th? N:^ liere \s that each nouri 
or fiDQii phrase marked for a j!^ ^ en caEe will be oji nolal^ with IbeEaine gramm^tjf^aJ 
fuimticih- "I'IiuJt Lhe p<H5Un>n of tln^ ncmn nr no^m phr^i-^ drn>- roL maclf^n o^lj iLj 
Eiafe-DiarkiiLg \s Involved in the d^^ermin^liaD cl ilfi grammatiLH] role. 

(4) &. (SUB J) 

(CASE ^ EE<J1 
b (OM) 

[CASK = AbS) 

Anni>U-tfd r-etrturtiLM-cletnutita art mapped Intcn f-stmrtijjp by thoir gramniat]- 
cal fuDctLoi]. ThQ f-struc lure achema for transitive v^rbb ]e sho^^nin Iigur4> 4.lr TIlc 
^PHEft^ fi|ot£ tn theJ^tryctiirn* Aif- nlj:k| tci hold lhi^ L^KKal itemft that cchrr^pond to 
Kt^tl]tn^t]^ funnztKLtis Lni]]c&tei] by tlie ^5UDJ and 'OBJ' sbt«. Likp^ llLC l^-XJcaE 
theory uEed b^ lIih^ pitr^pf, LVG al^t^ \hM th^ MhbfAt^^oi-i^ij^tlcn for aT^i]i]e]itA la 
dictaU-^f hy tin? prwiir.Ltor, (Ihifi ]Tifcjrin*Umi it sli-^w^n in tlir: tijp-kkv] 'PliED^ *lnt- 
Ln passings note Lti^t LVG LrLL^gr^U^hTTiatlcal fkjriCLmnf: Lo be ^ teint" n Uiry and so Lhe 
fijbcatn^uFLZatbtl h slated Lh Lbot^ tcrmsi GB t^ni Iheir^' Eitlkrtlohii to b« dmv^tlve 
from flomajitu: KJeWn and so fttatns prtdicfctlve^rleclJon in t«in citfl-roles- 



SdBJ 



OBJ 



CASE EltG 

CASE ABS 
PRED ^ 



PRED I LI ( SUHJ, OBJ ) I 



F]^r« i.l: ^'he f-strueLure Jichemdi for transitive Verbs. 

T^us we (DC Lhitt LFC functions v^fj Eimilarly to the GB an^fxiiint proposf^d in 
\h\i{ th^'i^ifi. Theseparalii^n ol prec^-d^nLh' aihd lihi'Tarchy is the key to b^tndliniC bnith 



apT'^U hn "^vzhetii^^cl^ iiuiILkiii, jdLi^Ix Dpr^Lhnif^ 



IQl 



free- and liiwi-ordtr plan^ticfln-^a. Whll<- tlln^O ire OttiOf t heorttlivJ cllfEcrtJkfM libit 
I^M'Vfnt A merfjini^ d( the i^^o lin^uiElK rAmps, both LFG aitiJ GD ^Eiem to be gd 
the Ei^rnc Fnolidif \\tn\ li ja unloriijiifiif 1 hftt tK^r? ha* bfcr no publish*^ ivork on 



4^1+3 Xree-AdjuliiiiLg (rtanutLar 

A ftceJlt d-^velopmenL lei the framework of Troe-AdjaEiLnR Gram m arf Joe] {TAG} 
presents in ]nrcrrtitiiift ^-ZOiiilt of Iree-wo'd flrdfj- pli^Dnmona- TA1i[LD/LP) pro- 
v]i|<!$ Sorot dofTvi^Glion sl-rncliires over whirh Jirtcnr jPrrrc^riHT tt1.lL Ijo diiliuwlr LD 
ftlnjctureft till he LltQtLght nl AA ID mLei? that may be xnorc than one level deep, ED 
TLilf-R are equivalerl lo LD Btmclurcs l>td^^[^lh OiLe. For tKiJn|^ltf tlie eqiJ i "^ILi^^rtt of 
tJLC ID fulft In (I) Jli^ shown in figure -1 '1. TAC(LD/LP) ccienda iIue nolion by 
eU anting structure of irbiLiary depLb. 



NP, VP 



vp 



MP. 



Figiue -1.2. The tlmctnial equivalents cf the ID nih^ in (IJ. 

A TAG(LD/LF) that n^veis the lingiiajt of trariEltJve sHfcteprw is ^bown m 
ti^re 4^{. Note that Lhe ^ammax cnrLndes onl> a single dftmlnfLl ion dLrntlerr, 
bnd no lieerLf preft^j^nM r<:litli>nfl. Thlp graiiijnar tloe?, indf^> generate the aIm 
permntaCkons of the laji^na^e^ Fuith^zrmore^ it repre&ecLtfi th? bi^janrllic^ rflatkncis 
that are lin^lfc^tirall^ Tnotiviht^. 

llowever, theio ia a problem with TAGlLD/LPl- rSir gfiwnm*r has too gr*^* 
an eKpieBstvt power There are no conslraants on the h &]npositicj[i r>f tbc clOrtLinii- 
tlo]L structuri^, &o srniclur^ of ajb]tfar7 ^zd can be ni^rd. Tbis frvrctom itllo^H 
TAG(LD/LP} tot»fn.]icllii- lirclit* of IrVl^F arid r^prraenl fiee word order J at least 
as far aG LbD Bnniple language dH^mHjhBtratrt), But TAO([jD/I.F) enJi^ins io h^ Ioq 
gdLioral^ and theieby Igac ilt- explanaLoTy po^er. Tor tbi-^ rrriM>a tbe GF( .anii^O^tnt 
prwoiilcd liflre si^nn in be prelf^r^-hriE. 

102 



HF VP 



V NP 

4-2 Shortcoiriin^a and Future Work 

PGrbapE the Tu<xl obV]cius ^hortconiin^ of Lh? c^iin^tLy j[[Lpk]nC]it4.^l |>aj^Cr ]s ttiL' 

ural l^n^agc nontAjns many LntracaLc: pbcDcun^Tia, and this parser bzE only bc|^bin 
Ln 1-nily ^[lillyse it. In Ihe IIt^I fi^^clion hclow I ]3£t fir^im? ONtjUrLi^itig COaiRtructin^tiiS 
that dcB^rvE attenLioD^ and dtscuEs Jlow the paiBcr mi^ht be xncdiFi^ Co h&ndlG 

Th^cth^r ^^w Ln vjtiirh the paiv^r i^gifigr ij]) i^lLael Is Lhg ^bilily to pane mni^ 
(ban oi]'.' lan^ijag?. GB i:^^ aft€r all, a LbcOry ihE^t atL^mt^ls !■> uXpL^Lti UiLii/rf^ 
f^raniDiar^ and so a parser bucd on h HhouLd b? ahile To par^c mikJiy lan^a^^^ nat 
Jij&l OnE]. Tb? Eurond F^'clLon dificiifciS4]£ wljfCt sotECLf Lo lie in^i^v^ in ^Xt^nnljibg tlhC 
parser to covei other LaT^^ua^es 

4,2,1 More Wc-^rlpiri 

TliWJ^jfl mticti moni lo W^rlpiTL Ihan tlkc [^l-:turt pn^insrttd Iji this tSiesLq ^ Tabi? i3 
^Lvcs a li^t orBome of the reiriauniD^ phenam^na lor Ihe parser Tbo:^ ]h\^r\ in the 
lir^t jroijp &re tjye nnjut lJkfl> to be L^^^L-^Jt^iJ wJl.Li JfJfttlvely IHtle elTortn The second 
f^roup shows a serious Bhortnom i ng of t he parkier that nknsi he addressed before Ibe 
partflr -rajk he *iMd to proj^H^tij" ijarnt lutura] Lan^uajser Tlit l^at gropp lists oth^r 
phenonieiia that wllJ demand a m^n-trlviaJ amount \yf woik, boih linguistic and 
ronipiilaiionaL However, even for these phenomena, the required modiliLatiO][a td 
thi^ par^i^r should f<J]oiv In tht samo vei]^ a^ thii furrcnlly Impkiriented systemt no 
niijor rework uE il^e ha^it engine sf^nii rft"«.sarv- 

Perhaps the inoet tantalizing phenomenon is parsing rootinHou^ ^^^ |)hf!^3<4 
^itb Inlr^j-mn^liitH- t^-^e-rtiarting Thl* h ffttiaonttrft-rj^l In |r>}, jepeated from the 
lntro^liji:iL>ry ch*^pt<T. Tin; fast jjJtr^*^ rLHhsiets iff tiivo simpler ra^e phrasesi mariu- 
(uand pv^up'/pa wtrt-Jji, both of which ezcut in a single phonoloj^acaLphra^?. Haildliti^ 
Likib p tieDon^eNO n Rhoihid Involve no ttl^^' tb;irt ^|vj|hg ^'JiSh'-jnarkers the plink'^aLnrli^in 



Set; [NiLf^] fiiir Eui cxE^nmT? diECLuaion oJ Lkr izJicoomm^ iiif Wntlttn. 

Ifll 



Q CQBttrtiiTilS i'Ke ptita-SC- with ItHcrmcdiaXcriic-marking 

a iJEBCcnljiiiiniha ca£c j^hrofics 

iLDadEliip in mitltl-nciqn <^flJe pltn^H 

& kxk^ unbtguHy 

& strurltiral i*mbl6ii]ty 

■a prtvcrb* 

□ lopk^ljialion 

Ldtinitjval i:l.iU£M 

d secondAiy pTcdicatlon 



Tnile ■!_?■ .Some jqtiATid|c>d WarlpliJ pltf-notm^na. 

oTselGctinfi an u]bhQii]tded numhororoth^TH^AW plin^es. Willi llkis ii^tin^n, fuch CMO 
ptnftt: woLllJ b^ pari-4>d by EirfL p^rsis^ ciclh of tbc^impl^T cDnLLnuDua ca^e phroEGs 
(4^ these witli EGiDC number gJ un marled naun^: fcplln^^wfiJ by n ^.aa^ ni£-rk^^ Lln^iimJ. 
Wd th4>Ji grcaupinE tllc^r^ iTfkfi^ plir^:^^ V^ltk llie aewl> ^d«l Bclion. 



(5) mar^N-tu pN^wripa win-tt 

largajoo-UAT frlpnrtlj blg-&AT 
^to/ffOiU llhG big. friendly ka]L|;aroo^ 



PaniJIl^ [L]&COQt[ilui>i]A f^atf plird:^^ alfo seeing to be cleee aL hand, (ti} ahaws & 
\^riaLioTi nftliefra-mph^e^^rLlenccffOia tJitirLlmductiH™ wiLLpucb Q-pbrase, campceed 
cf tll^ i^'Urd^ itu nf £^tu an^] ifFrf-^T They arc inLerprcLed in unison^ roughly ob 'the 
bjg child V despite che £?p&rat|c3r Tr th^sj^ntcncer ThtactLon (o he iddfJ litn! would 
nhow up it] Ayill^iX^ ^ a£ tu L|E^[in] tbe eF-?c(£ of ordering Rouj^bly Epe^kiDg, ca^e- 
marki^tii WuLild b€ allo^'Dd to lake F^imilar]^ rnitrl^ i^aj^^ phr:iJie$ ^ argumi^nt4- Rllt 
the prfH^]-^e uLtiKtnn^ to IjC Jr^rl^^ ^t*^ not soJiL^r ^hen tbe ai]i]yzii]^ the ln?3d of 
tll^ [nEir^^ ]S COiLSidcnrd L Tiiat iKh i^ne noun of a caae phrafe ]& iDterpTetE^ a£ the 
hj?ad or tlie phrases with thfl other nounn ^rtlng »* inoilillors (ttiuch llki? a^ljoftlvi?! 
ill Engb^h). Tb? duierriLLnation l^ tbe b^Ad i>f a phrvin^ ajarl ihn^ n^c^lilic^LiorL^I 
Etructuredoea notgefm to bf SO fl UAlftb tfOrW^mJ , yn^t llieineji^lLe]Ltt (orpATyinE rull 
caac phraBJ^ st^eph ta lie ftv-tlLlihlfl. All Ihaft in Heeded ia the Ihwietkal reripe_ 



1-E]na iHrp&r-lfi-3cl uit^^ povpa^t diilri cat 

bd07nt>nui§, big cat iq^ 

^lam tukinfi Lhe bonm^ranft from the hiy clllld/ 



As menlioned itbovfl, iheie is inoLturr -cltltn-nt Ln the iUKiliary Word^ ilMiHilj'T 
torfLplementizeiB- ThpiC Hettn^ntfi ludkate the LCLocd uf tEiC ^M^Llteru:e &i]d cojiibinE 
Willi Xh^ tf-nflH? Mid ftspecl Lnfcir]jiailiniL Murphi^loKicn^lly FjJi'riUfLft, they appeir 3ff 
the iLrEt efenieiil In Mie hrn^ar templatG o( thJ? M3Klll--ir> bo this forcn of patHliift 
^Dil^l requne too cnurh uJTorL SynLMtiiZii:^ speaking IIK7 iJtjn'l preaeiH mncb oEa 
problem eithpin as lliey i^ah be entered L]bta Xt\A fLCLKiLtary projetUon. 

Tbe next &.TCR on the Ittt tt the tradltioaal problnini of ambiRaLt? ^hiclL Iht 
current Vei^Lon of Hie parR4ir assum*^ .^way- There are L^^ro (li'-orh dl amhieuitj 
tli^l f*ll under Lhe purvir-^v ri[ Lhi? parfier. LeXJCaL iJT]ljlgnLty euBLG when A- fiillsV 
morpbeme rnipft to frlOJ^ ihap one Itxlcil tJltry, due to dirTcring llSf^ In ayittazc 
Syntactic imblftulty occur*. "cvhr^El nin^Tr^ than one tlrui^lnire rn^y be deriMsd hi ttnf 
jiare^ of A singre BeTit^n« At pieRent [ cnii uHi^r ilO bptter aoCutlon otlLCr thwi L|lp 
BLaud^ird nn^thnxl-^T auch a^ sLmulatiOii oE * Tion-deterministic |i^J^Pt or insLallation 
ot a lookahcad di^^jnl* td remove Local aiubagaityr This area orpikni][iB dt*^rv^ morE 
atteotionL 

Morpholtj^caL arabiEuHyr *he ^^^^ ^^P^ liBted, ^^lll arlne when the pitscr Is 
extended Lo handle uJiStp^ralfMi wordB. Tiii^t Is^ instead of Bupplying htiltfltliMS wilh 
Mlit^[H^s phraaei word^ and QiOiVlLCjme boandarLefiH only Llni former three will Retd 
to heffkipplied The job ofth^ pfl-TSn^r then ejpanJs to hin^kins np each ^^orJ intoitu 
consLitupiL lliorph^niefl For exam]ilL\ thfl wnTcl n^ajEJ /n rJy ^_^^lk he broken into two 
rnorpftftnP Miwrs, ahown in [T). (Ah iL liappenfi, Lhe [ir^U tfVfr Is ungraminatical, ifld 
only Lhe Btcond— the one used ]n tho WJnple ae^ Lence -is graccimaLlc^Jil ) Ambiguity 
ujsts when more Lhan one tflvtj- p pcsBlbk. auch a£ tlit t^^e ^ven here. Agwn, 
one BoiuLiOJi that ropfifta lo mind is Mi trj" earh fover Beparately (trC= , hy iiimuliLinR 
t non-dEtemiLnj6lir pitScOr and hAltlng the pansi: on thH>W f-ovej^ LLaL do J]Ot pan 
opt for some re*tiOfl (r^rg,, fAllIng tororabinO itlorphCplogicaUyj 

(7) ^. (njMjy^ Jn, Wuf 
b. (n^jufu, rJu) 

The lait £rH>up In tlie Hat indicates the wide f^nfle ol conBtrucLkme m W&rLpirf. 
Preverbfi cllticize onto vorh gtema and affect tht dl&*ninft and aubcatc£on/*t|on of 
tha prn^JltaLor- For exaniplef tin: previ^J-b, ^^^t, adds a SjCtiefiCti™ aspect I0 Lhp 
Terb. NoLh^ thrtt th*^ rhiURe Ln meaning liMitSfHJ another argnin^ilL, which would 
alBO appear w]th dat]ve c&riL-- m n rkl ng- (3) Bhowh Llm Minple flenU?u^:^> fiortL the 
IntrtdUHlUcm with the pre^erb addn^ rr> th-P main verbn ajid tho e^tra arpiment for 
w]||ch it subcaLe^orize*. If fireverba were this airnplen It wouldn't be ti>odlflirult to 
rnikis the extensions for handhng tV-m However, prcVi]rb:i enter into miny Other 
con-Rtructions thitL arc 111- understood 

I-LRCE Li^PEILF-ls-^M-^^A^btllBf^Ctlve-tabE-NOUPASTclnlcL-&A^ 

boomerang man uai 

'I aJn tiJiLJig the ho^>mnrajiR frnm tliiS child for (the benefit of} lhe raeo/ 

ConipH>u tiding Is ^ morpholo^cal phenucnrnoih where two WOrds fom bi ne to form 
a i^lnnln^ woitl, i^iih acorrcFipondinfi I'liange m meaning. For OMftJls ple^ nnuna ^gr^^^ 



ajid unTHirnu ^eat' can b^ cnni pou ndcKl t^ form fho ma-npamrJ [[Lu^hlh^ A £t^i$5-c^at^r- 
Tlni tnorpholn^' rif thn^p worrlfr |b Hlrrill^r rjn tliAt of casp-majliinj;, yei there an? 
dincrecic^ whjrii wLII cnlaiL rj^cdiFi^Hlioii^ to tlio pir^er. 

'I'opLr.iljTaiioii is ths prnccss cf ultcrini^a pbrnn^ bnfur^ IhcMPtitt'ilfc^ ofLOii ^ith 
a ELfijilitaiLl pai[St ill bcLwcoii Lihc Ffrmr^ p]Lrt>it^ and tNe Mnlciice piopfj, Tot t^|e 
puT]]Hist of tiuphftals, lliis 1? ft srrongej- fortii rliaii mnvinR a. phrase Lo the Brat 
posit]ci]L_ In facts there caJi be se'-praL LopicaLLLSenl [HtirMti:, Fhlftti^rmon^T tL>|ilcal- 
lltitliMi will tiaually luvtJvc rt^J^tltlon of p-hR^es riLher than movement^ and ao the 
copicKl phriEE^ miut bs ideEirified dunnj; jianin^ TbuE if cx^mpliEiH^ i^llh anO^h^'f 
ViTtation of llie saniplc- scrttnCO, slic^Wci hi r^)r The synl^iK nt topirq nJoea m^L BMm 
tiXi h^H to i^Latf, bnL Lh^ real prohlem Ecea in its EecnanEjUH Thu th^urf CjT focuf 
and emphasis is sliJI quiLe impoV^i fc:hcd . 

(9) .^poj'uiu- rJUf n^_?Nfu-r/u id-rno-r/a jpUJiCu-rr+i turt^u-tu tdF'Jt. 

^As for Hi^k I ani taking Ibi? hKhome riUig [rtnti iLc thUd^^ 

Inlinitival clausBs, like in Fn^liEh, are BDbord]i]ate cLauseb that mntuti a v^rbaJ 
element and arl^umen ti:_ OAcn one of ttieae ar^ucn^cLt Lb Unkj^d tOOJii" t IlO argtanitn t5 
In tlie main phr*^tftr |10) ^hows an ty^rapt^ of an irfinitival clanHe (in thia cftfiE^ 
EontreilEed by the ^iibj?i.t of the mnin ciauf^)^ titk^jn Ixom t^0 diartLiSSlon of control 
tbto^> in Ihp sci-Hjiid chipter. HoVi tfiorr- ie no Jflultl- level i^inhwIOireflfsiibor^lLiiaLe 
Llansiis ip WarlpLri; inlinltaval clauEet inaT be nested joft crnce in a teHt^^i^LN TIL4 
aynliJ^of Ltj«G tl&ijPMEwniB viji^-sjmjkr tribal ^ImwJKlftiJS^^, ":tn<l ^^^ it s^t^uldp't 
be taa dilT^cnlt taexLend the parEeiL Perhaps the trickiest part of thri plLeitOirieilUL 
u« tho i3ijiic« or contrr?! ivlLtch u^ll[ rtH^uit^ inodifitalion. 

(10) H'^iYilT} tj]-ju t^>anrijt4j-f?ad ^ar^a ttirTo-Yija-kfirra. 
WODlAn IMrEHF-lo Eilt:Jk-SD^PA5T yicn diE-l^F-CO^^^' 
^The ^'oman If tpeaLinj; to me while d]f;|^n^ yams." 

S^ecc^^ary predlraLos are bke res trie El ve rdative claiiaeE in Englisb which mcdify 
one of the main argnini?ntt» of ^ ventoar^. Ah i:me might C:>lp^t, tti^C n^lan^^ are 
idenlilied with the ar^menl the> Tnodify ~by ca5e-markiEig{ratliet IhaXk by p[>sltlaiL, 
^ whtti F^ngli-ih)- In [li)i the senZGnfTajy Jilau^e mcrfliEies the Eub^ecL^ Lndicatioj^ the 
body pnirt bj TVhich thni attif>t] wa* ftc-Iiimrticl. Thn Byiitix ofscfi^ndary pn^kjitfs is 
like that of Inlinitiv^ clanaeE in that L]kere Is an ^r^ment ouLflide the claiue that 
l£ the Eubjoct of the cLiuso Lts-^lf. Aa i rn^^ulr, secondity prtJltrntion should follnw 
fajrly (|n]ck]y ficmi the extenBions needed for inlimLi^ clauses. 

(11) /Vpffjitiu-rfu irt-rnft-rfd punJa-mF rtiTcJM-rtir itJnTa-tw tflf/r 
l-l:lLf> lUPE^F-ts-Hd take-SO^FA£T- hand-Bbf^ child-DAT 
IxKHnerajig 

4 am lakiiij; the liot^ECLnTang from thv rblld witti rtiy Itiind^/ 

The East pheDomoeon ^j]h the Ust, noTninH'J smtenrt^. i[< ratb^^r I'l^ttLlhCin In Irui- 
glr*fies of ih^ i^orlrf- Nomiit*] aanlencia aip oft^n u^ed to £laLe a feature about 

100 



BomalLing. Thty fAtitiOl tui used to dlsc^lB^ bCtlors, Ibl knfH^n<^er m ^oiqU be done 
with verbal aentences. [I'l] Rtv^s an example of ^ noinin^l freilteilcC- Dcfplt^ t^Le 
Lack ci[ a vctIjt tluft wnterM c^n bi^ mterpretwir ard It Is given the readiDR Lliat a 
CGpul^T provides in lan^ua-^es like E^glt&h. As WElh tome dF Ihe dllhcr plicnnnkLinlL 
ll^tHMl abi>vi^, thn? fiinipkr r^^9 fif m^rriiniil spnteTiren don'L ^eem too far afl~ for Llii? 
pFLraer. Dut once t|ie real complexity cE tbete freiit^mxa ifl chcod lit^^rcd, tlLCir tutry 
itllo tllC JJAJ^t's rtin^rtotrf heronifts a. hit more dtatajiL 

(12) JVffdrrid-jm-m-pJn «r,"Ti-jrttrrt 
nian-UUAl-^aft blg-ntAL 
^The \^c niQD are higr' 

4.^.2 Other Laiiguag^A 

With only on* Iw^gnag^ In t|i^ pifHer'ji doinain. it i^ rather easy lo chcote amOng 
theremaLncaR LariE^a^s o(iht WfOrld for Othnirt to bi: j^ArswI. Ttil^ thnislg has liiDted 
^t huw En^lLsls [[Light be |iarEJJd, 5.rtd t^^t^ Fn ISct, Is llic nt^t projert T Li^tend Lo na- 
il ertahe with Lhe parser Happily^ jL Eeems (hat onFy a small amoanl or[[LOdill ration 
wLLt be needed. 

The key to tail[)rLng the parser to Euj^lish [s to reco^ize the cOrr4.-spO]^d(n{:CfV 
bclween ihe W^lpin phEoomena and Ih^iJ- Engli^li COiitltorpart^. PinL, consider 
RLDrphoLogy. VerhaL morpholo|[:> m EngLi^h Eeenis to be 4i6 ^knplr] ^ Waf[j)itl. wltji 
ih^ tense elem^^nt ftfli^jng onto th^ verb atejii_ Of couiEej there are maay more 
^-:^4>|>tlc9ri« with English VH_-tlje, but tht^ic can be d^t with later- 

fia^ifh nominal [ciorphoLogy is simpler because Ihcrn: an: nu cati^-in^fken (dlfi- 
regerdLni^ the gariLllve case, for InsLanccl Thus^ only aouiif lUid numlin?!- miirlierfl 
[ir^.p "^-t") will fiAVii tH^ be coverwlr 

Dative C4*Ee-[iiirkin^ in English it pH?riormcd at the word kvel and troni left-to- 
rji^ht, a& opp[K«] Lo the morpholop;k al LevGl and from ri^h I- lo-Jefir ThiH difcltLil<~lioiL 
will h^^ iLuiplt to tJ^TWk in tJie lexicon (The jiTtpoaltiona "^to** or -^rom"* are the 
Tn^jn dative caiie-markcT? m EngliEh.) 

Nommativecase-markLnKiE perEormE^ by INELflhc EngliBh ei|U]v^^nt of AUX], 
find ah^olutlve r^-^f - m :^rkl n g ii^ pfTform^ by the verb The verh follows suit iti lU 
caje-markin^ direction of lelt- to-right, hut I NFL a^ems to mark it^ ca^C in l]\t OthLT 
VfiVr This secrtLin^ fuiitradJttiOrt J^ aLtt^dy EiaJldled hy the para^^r, aa \i processes 
th^ an x"]! I ary ] n especial manner anyhow In EogUsh, the auuliary wLEL jsimply i[t>xth 
Its c^k^ t-Eght^li>leri^ rathtr thcia Icft-to-righi ^ hk W^rlplrln 

Al Ibr ihe syntactic structures, they wLU be very Eiinilar to those ^rfed fot Wwl- 
pirl, as m^rittoned kr> the iji trod n^^tory chapter- One discrepancy concerns the p^- 
sltlnMi ol the flubj^jtt. In the WirLpLn GB [itcraturen tlie gubiput lh placed Ln the 
spccifiH_"r position of the v**rhal projection, whcrCf*s In ihe mAlnstrtam fSB theory 
{which hab meet olten focused on Englisb]^ the subject is placed in tke ^pcciAer 
pofibtlort oUNFL 1[\*' |iiLJ^er B Ellr+*;uly powerful ennujli to encode the dis^nLttrjn, 
btil the tlLi9ot?tical diUGnincGE shoiJil be irodcd out in that 4 more unilied structure 
can be n^ed. 

107 



Thlfr la only h very mugli &leLcti of how lo bnilJ a iTorr^pondlng paj^r Tor 
E>l|B|iBbT but jLFlhriijIrl fn'tvr Lf> | flh] L^ Fitn^ thr ff^lati^o BiuipLLrity oFll^c I^Ll. Ct Cr^tune, 
il rtSLiinrv to be don^^ buL ths i^Eknei^ loohs bb« II v^iLL piove robubt ^nou^h Ibr Xh^ 
jnb. 



log 



Appendix A 
Test Cases 

A*l Implemtutiitiou Not^js 

JuhL a couple oF iii7t<* fttnmit the 1 rnpleirLPiltitl&tl . FlFHt. 1 flbould raEmHon tJial ttifl 
paTE^r te Ktmlly quite laet. It tllkCE ^hoiit dii€ Ui lwo£CC<Htidft to paree Lk^ sample 
&erHeiitc, itld rLOtniucli liHiRer tointtrpit-t the reau Itiiift ay d tactit fc tfuCtU re to obtain 
the tf-ToLe/wcrd mapping. 

Tbe codn? LtfiOlElft ithoiJt -W papea iis Vitglli. There are 19 objects tliat fomprti* 
the prrj^I^m- Tbe rnajor objwLfi an? ttie precftieiice pirfier*mJ tite fl.}[iCactLi: parBer^ 
as well M thtlH^Utnn The preced eacis plLt*(]r iwn9l&ti of lour objett* for ta^t of the 
1ihoiioli>&ca|lgvd5.andonDCPrtU*Jpirs*rconlaimngth<:ba3ti^en^^ The sjnVai: tier 
pai^r Es A Siriftk object- Itoth pAi*iKa aif baaed oi. tht phnu:4>m;irkor object that 
linplemfiiLs astinpl^ forest fttmct-ni^i the precedence jiatStrlmpcRcs ordering on (llC 
lorevt, h"jhi^reai the synLactii: pairer dots Hot_ The phrase- rrutfte J", IJi tufiLn refei* to 
jatCftOfy nhjecta that impleniflnt tht r<rtl-es of the parse ire^. CfLtJ?^orifts contain tho 
data and a<_tlonS partimilar to the parser typt Tlie levicon^ on Ihv hj11]<^J' li^nd, a 
fOr-=:tnirted a£ a mappilLg of morphemeE Lolexjcal cntrlft, thi^Pruelvea objects in the 
Byst<im. At the base of Lhe system arn: ftve support objects Lin]jlertitn Un ft ItfltaTaclaj 
mappinfis, fuDctioDS And ^rray^i 

Tht piogram was wrilttrt in the McFLavor By&tcm Oil * SynibotcB StM-BH^rlM 
Lisp Machine funrtflr release fi_l)_ The Hfflavor flyetfm Ls aji tibjeCt-ortented flavor 
BystemT^rittfiiM MIT by Edward Jl^jUn^ which ib very mucMtke the Utp Macliine 
ll?LVor Byatem. McRii-W w^S d^osen because H rndS ill Maclisp (nrderTOFS20]r n-^ 
well as Lite Symholltfl- 

A.2 Testa Cissca 

Thin BectLon contiiii^ i {ruther lorj] BcrLES of twt CWes for th€ paJiier, Tlie tr^t 
tjijet: 5ire listed id table A. 1. Rnr eafh type, a number of tHt* weiecoudncted, and 

they are lifted In thtlr rorrespoTadinE heclioJi, I'rgrammatical inpictft flie Irthelled 
lylth An^tensl, note tliaL all of t-liem have, indeed. b*^ii (Ltrlared ill- formed by tjie 
pafHtJ-. Grammatical inputs are pn-Stnti^ without annotatl^nT and notn that thr 
paiscr hat properly pn'ari^sgH^ them tfio. 

109 



a vertiB 

n noun i:i>iU|H>SLliDrL 

□ verb i:o]ii|>i>ftfi|nti 

ci auKiliarV cotnpc^itlotl 

□ cr>[itj]iiirtu^ H^a_^^ plir^efl 
■3 auxiJiary pofiirlnDLii^ 

□ frn^r: plifaw ofder 

D ar^m^Eil ]d4>nlklic^Liotl 

n null auxjIiATy ixHinpuilci] Ul 

□ tiull ftnapliora 

□ loo [rtajNy ^4^Tj]]i(MlU 

□ ca^^ niartin[^ 

Q aui^iLiary b&ue ij;rc4MiieDL 

TicmJiiaC ^^rttiiucnt 

T^ble AJ^ Test 



A.2.1 Verb Stems 
Paxslng TIl 

PS: 0^ data, cn N J UQ ATID tf ^ 5 
EKirpheiDQL YA 

catfl£C^rJ: VERB 

5St prpj^ctlbft?: HTL 
Cftt*joTy VE3tB 
children, prtijictloti?- T 



actLtjriEiL TEETA-liSSlCrHr THEHR 
catflgory. VERB 
thildren: proiectioa7: T 

actLcsriBr SPKIFIEB: (C4SE-iS3ICtf 

l,B50LliTiVf) 
diti; THETA-ROLES : (THEHE) 
THEME- AE^DLUTITE 
SUBJECT. THEflE 
Bflrphana ■ YA 
categDiy: VEP3 



Paremg YULKA- 

PS: Oj data: COBJIKIATIOB; 1 



110 



category: VERB 

SS; prcjectii]i]Ti NIL 
Cartgory] VERB 
children: projactio:;?. T 

aC^ic-nEL TKETA-iSSJCB: THCT£ 
catflgi^rj^ VERB 
chlldFen: projflttion?: T 

actions^ THET*-LICnt. PATH 

SPECrFIEft: iClEE-ASSrcS 
JIB5IJL-UTIVE) 
data. THCTA-ROLESr CTHEHE PATF) 
THEME: ABSDHJTIVt 
PATK: DinvE 
SUBJECT- TBEHE 
OBJECT. flLTB 

catB^ry- VEflB 

I'AFHi.Dg VAEini. 

par 0: data^ CDS J UG iTlD Ef ; 2 
nrorpilChnB: UAFLRI 
catBfiPtJ: VERB 

ES. pri] action'^ ^ MIL 
catagtjry: VEBB 

iCtLchCiBE CASE-ASSIGN" ER^jATIVE 

TKETA-AEEICkN; Ai^EUT 
Cttfigory: VEflB 
cbildr«n; pr&jactinn'^: T 

aCllonfl" THETIi-LJBhi PIlTB 
dftta] n^Etli-TLPLES: UGENT PATH) 
iCEHT; ERQATIVE 
FiTli: DATIVE 
SUBJECT. |il3EHT 
OBJECT^ PATH 

f:at*ff?ry: VERB 

Parflins ffTl. 

P3: d- data[ CQNJUGATIDH] 3 

IDQrphfl^*: WA 



in 



SS" pro j &ct Li>n7 r KIL 
rl^igoryi VERB 

actLcjciD; CASE-i^lGH- ERCJiTrVE 

catflgc^rr^ VERB 
clkildron: proj+ction? : T 

actlonB; TflETX ■ ASS 1^1? ■ THEKE 

CDHPLEHENTr ({^ASE-A5SIGEf . 
ABSULUTlVEf 
data: Tl^ET Jl-HD LC5 : (AGENT THEHE) 
iGESTr PEltilTIVE 
THEME r ABSOLtTTItfE 
SUBJECT: *aH(T 
DEJECT: THEJffi 

tat^gury: VERB 

Ptrtfing PlTWTl. 

PS-j 0- d^t^: CQHJUGATIDEf: 2 
morph^mn: PUNTi 
citfigory: VERB 

ES' prelection?: NIL 

childrfln: prr^JwCtir^n'^ T 

acttoriE- CASE-ASSIGB: EBC/iTIVE 

THETA-ASSLGEf- AGEHT 
cansgqry- VERB 
children L projection? ^ T 

ac-ti&M: TKETA-^LIMK: PATH 

TEETA-ASSIEtN" THQIE 
CDWLEMEJIT] ( CItS E-A53 Lv^H 4- 
1BSQLUTI"VE) 
d&ta- THETA-ROLESr U^^KT THWE PATH) 

l^cEJ^■i eucjlTive 

TflEflE: ABSDLUTIVE 
PliTll^ PATIVE 
SU&JECTi ArJEMT 
OBJECT: PATH 

category: "^EflB 



US 



A^3.3 IjiIIdcIliI Verbs 

P5: Q] catagorTt VERB 

tMldrM- 0: data^ COHIlWfcrTflll S 

BDrphQDL€1 Ti 

ta1;4gDrr^ VQU 
1: dlU^ CDNJ[K]iTIDff[ B 

5S-: pTiSjftCtioFi?: NIL 
cai:agcT7 : VERB 
childran. prci]4Cticn?= T 

■CtlOM= C ASE- ASSIGN ^ ilB&DMrT]V£ 

THETl-ASSrC3ir: "DIEKE 
citftgory: VERB 
cblldrwi^ pr&jaction?: T 

data: TEKSC: NDKC^lST 

THETA-RDLES: <THEKti 
-™ef4E J.B^DLUTIVE 
subject^ tbeke 

Cit^flory' VE3tB 

FAFBing (yULKA HI? ^ 

PSl [>[ categcry. VERt 

Chlldran- 0: data: CDNJUCITIDN. 1 
QvurphfliDQr YULKA 
^atagorr^ VERB 

1: dAta- CDPJJLVJATIDEf: 1 

DiArfhamfl t HI 
cat&Eory: TEHSE 

S3: projection"^: IrfIL 
cate J&T7 ■ VERB 
Cbtlflr^ft" prciiactiQn7i T 

actions: CA?E-AS3iaN: iBSQLlmVE 

THET,l-A33rGJf^ THEHE; 
eit»iory: VEftB 
cblldrao] prnjaDtlaa?^ T 

ACi-^cnaJ THETi-LIEfK: PJlH 



113 



dHtn: TOHE^ H0SPA5T 

TT1ETI.-3UJL& ■ [THEUE PftTHl 

THEHE: ABSDLmi^ 

PiTHT DlTlVE 

SmJECT: TldEKE 

□EJECT: PITH 
marpkiaBfi: YITLKA 
catagi^ry VERB 

Pftraine (VAItfl[ lUill- 

PSt 0: catQ£Ciry= VEJtfl 

childifln: 0: datli: CaHJL"{]lTI[JNt 2 

mi^rpheniQ: VAf^ 

catag&rj- VEflH 

l: dfttai COB-TJCATrOB^ 2 
catflat^rj" TESSE 
S3: prcpJactLflTi*' IflL 

THETA-liSSlIN- AGENT 
cataaor?: VERB 
dalldl^n: prflj^ntion?^ T 

Bttionj: MCEffSE: < C^AS E- ASSiaH - ^BSOLUTlVEl 

TflETi-LIEnt. PATH 
ditl: TEM3E: NOBFAST 

-niETJ-'ROLES. CWSEST FlTHl 

ACEHT: EKQATIVE 

PATH" DtnVE 

SUBJECT; ACEWT 

□PJECT- PATH 
nnrphana: WARAI 
categorj- VERB 

Parsing CNVl ETfl), 

PSt 0: catagary: VE^ 

childxem ^t data {'{]BJtraiTIDI]: 3 
mnrphflJM' EfTA 
catBfii^rj- VERB 



114 



1: data] CDV JIJGIiTIDll : 3 

ncTphana ' NYI 

3S: pro J action^. ¥IU 
category] VEBB 

actici]iB: CiSE-iSSIGIi; EBCATIVE 

THETli-flSStatf: AGENT 
C4t4£0Ty^ VERB 
childi^D] ptijJ^ctlonT : T 



actlCinfl^ Ci££-ASSrGJI: AESOLUTITE 

THETA- ASSIGN. THKHH 
dita; TEFfSE; XDHPAST 

TE{£TA-ROLES : CJlGEXT' PlEHE) 

AGEFTT: EJtGJ.TTVE 

THEHE; IBSOLUTEVE 

SUBJECT: AGE3IT 

OBJ EOT: r«EHE 
i&DrphQQie^ NYA 
C4t«££iry- VERB 



PirsIciE fPUlfTA KItl) h 



PE: 0: citagnry VERB 

Childran. 0" d*t»: CD N J JC ATT DM : 5 
Dflrphenfl ■ PUBTA 

i: di«: caHJu&ATinrfi j 

category; VERB 
chjtdrftiii pro J fliiticin? ] T 



*«lcnft^ ClSE-ASSIGH: EJIGATIVE 

IBETL-ASSIGIT: AGENT 
CAt4A^ryi TERB 
Childrflii; pHj]flCtlfin? ^ T 

tCTUiTi^i CABE-ASSIGH] ABSDLUTI^ 
rUETl-LIBX: PITB 
THETi-lSSrCIf; THEPTE 
data- TEVB^. NDHPJiST 

THETA-RQLE: (ACEETT THE3l£ PliTHJ 



llfi 



ACEPTT- mCATIVE 
THEHE- AE50LTJTTVE 
PiTB: DITLVE 
SUBJECT: ACER? 
DBJECTi PATH 
ftarjriiBiDBr FL7NTA 
CB»gprj^ VERB 

A-2-3 Ndud ConipDsi tiDn 

PS: til phrUAl ItCllvn*^ SELECT*; HDW 
CbtvgcHrr? CASE 
childreiLL Oi narfhdiafi- VE^XJUI^tJ 

ct.t4g&rr : cliSE 

SSj prQ] actioTL^ ■ NIL 
Cttegi^ry; CASE 
cbildren! prchj 4<itl&£t? ^ ITli 
datl: PetlSQETr 1 

irUHBEElr (SIHCIULiRt 
B)Drpfl4m4: HE^AJr^LU 

projgcticm?^: T 

data[ CiSE-HiAKED- ECU ATI VE 
marpA«i4: fUJJ 
catflgcirT- CJiSH 

Parainfl *(KlJtLI flfll>. 

CEARLI RECI] ia im^ruEOLELt icU - 

T7i4 precfidQDC^ Btr^ctur^ la imc Dnnac tad . 

Pirslbg *(HU BnmDtn- 

(KU HUHI^U) is ungrumiatLcal. 

Tha pracfidencfi atmctujfl ].p unconn^ctod r 

PaTsing 4(WiTtLU KU FXUJh 

(HARLtr KIT ALU) la mgranfflttlul h 

lie 



A. 2. 4 Verb Compo'iitiqn 

Purging (PUMTi MI). 

PS: 0. catfifiCMj: VERB 

cKlldronr 0^ data: COlfJUGATIOIf^ 2 

cat&^ory: VEFlB 

]: data: CDNJUUlTiaW: 2 
DLorpbeme : 731 
tatfrgory: TENSE 

S3: prQJBCtlQn^[ JfIL 

chiJdrftn] pro] action? [ T 

TElETA-lSSIM^ iGEBT 
CAtofciry. VEJtB 

actionH. CiSE-lESIGff: lESaLUnVE 
THETA-LI PCK. P*™ 
■raETi-ASarCEC: TMEME 
d4ta: T^EfSE: HOHPiST 

TWETA-FLCLZS: fACENT IHQlX PAIH) 

AGEFT:: ERG ATI VE 

THEME I ABSQLUTIVE 

PiTH] DATIVE 

EtfBJECT. AG EST 

DBJECTr PATH 
morphfriDQ ' PUUTA 

tXYA KI) id Ung^^lffl^]ati^al. 

Tho precsdance BtTuctujo In unconn^CTOd. 

Parking •fKU YULJU). 

Thri pr*ddd4hCv structure ib un^oiuie ctad ^ 



117 



A+^+S Aujciliftry Compurailjou 

(nifJi) is m]£myH:tlc&l. 

Th^ au^cili^fy h&^ tQQ f^T syllbbltB h 

[FlLIi] is ungr^Ntmnfttical. 

Hb ^axiT-laiTT has too fav pyLlflblBfl. 

ParBing •IMB RLl) . 

(RKA RLA) 1* imftrEmniatLcal. 
Th* irord bBginc vilh a clitic. 

Paiaina *(KA) ^ 

(KKi iB ujigramniati C^ - 

TTia auxiliary had too -few Bjllabltin 

P«raina CKA RliA), 

P3t 0: Ut4gcrr7- XtfllLIlRV-BASE 

childran. 0; lexical actinne : lUII LI ARY-S ELECT : 

iUII LI AHV- DATIVE 
AtlJILIJiRY-SELECTT 
tUIlLI A JIY- OBJECT 
data: SYLLIBLES . 1 

CBTejoTy: JlUII Lt 4fiY-BASE 

l: leiical oCtiOM ^ RIGETT-AD JACEflT : 

fAUlILlARV-DBJECT 
AUKlLIARV-nATIVEl 
data: STLLiieLES" 1 

category: AUJLHULAay-SUSJECT 

G3; proJfrctioEi^: MIL 

citafiorjl AlWlUiRY-BlSE 

ch Lldr^P : p^Q j 4c ^ lot^T i T 

iCtiDEiB: AJlClfflEETT: VEHB 
cotflg&ry: AyiiLiABT-aiiEE 
cbildr&n[ projactiont ; NIL 



Ctteg&rjr AlHI L! J, RT-.^^UH JECT 

priJActLon? r T 

dftti; T^XSEE: 1HP[fFil3Tl 

CAtS^ory: AUirLIftB'|-&A5£ 

PS: 0] c&tagar^r AtfULI iElY- 3 JiSE 

Childran: 0: Ifljric*! actions' AUXlLIinr-SELECT: 

AtllIUAFiY-OBJEi:T 
AtJJTLIinT-SELECT- 

jLUJiL[Afty'SUBj£<n- 

Borphflmt^^ KA 

CAtsfiory: ACJlILlABT-HlSE 

1: dati: 3TT-LJiBLES= 1 
BHurphA&B ALA 

C^tigflry: AUII LI A J^V-UATI VE 

SS] prD^*ctlOh7: BIL 

catB-^ryj il7XILlARY-BA5E 
childraq- pro J &c t i.oii» ■ T 

actlord. AEl^^UKENT- VUiB 

CBte^ry^ AUXlLIAaV-EAEE 

Chlldrao] projection?] HTL 

category: AUJI LI JlRY- DATIVE 

pre j act ion?: T 

dMi- -re»SES: (liOlfpAST) 

ASPECT. IMPEflFECT 
norphsDO] El 
cit#goryi AlirLTAFlY-BASE 

FBFHiT]^ (KA lUA RLA), 

PS: 0: citflgprrr AUIIU AFLT-SUB JECT 



lift 



AUtlLIlRY-UBJECT 
data: 5VLUBLE5= 1 

tatajtiT?: fiUilLIiltT-HlSE 

1: dfttn- STLLJiBLES-; 1 
murpbeiDQ . ilNA 
catBfiPry^ AtUILIiRT-aUBJECT 

2^ dara[ SYLLIlBLES:; I 

cata^ory: AUJ LU flElV -CATITE 



SS: projection': MIL 

chLldranr pr«Jaction7: T 

catBg&ry: AUIlLIAfLT-BJiSE 
chlldran! pr^n j act ii3Ti? : NlTi 



^^tajoTyT ITJII LEAHY-DATIVE 

pro] act ion? r MIL 

ECUnPEft: &ItfGULAFL 
borphaner RKJi 
catfrgtrj: A UlILl ART-SUBJECT 

pToJfictl[m7 : T 
data: TZK5E5" {BQHPiST) 
ISPECT: ITFERPECT 

iDDrphflB? . Kh 

catagt^ry: AUITLl ARY-BliSE 



f«cilIiS +(LP1 WCA EiLAj- 

(LPA aNA ITLA) la uBfiraimiatical, 
Th€ v^rd b4glrLfi vith a dllti<;- 

ParEij.De '<I^A K^^- 

(lUfA KA) ia imgramniatical- 

T^a prfcadvnc4 ptTu^tuxa i« utiCOim^^tad . 

Partiing *(Ki LPAl. 

120 



(KA LFA} IB ui]gramn]atical r 

TI34 pracederLCb atructuT* i« iinc^nnactsd . 

Thfi prBcedaQca Bt mature is uncoiinQi^ted r 

Jl.2.6 Continuous Ca£>G Phrases 

PftT^lug aVIFUmiJ!) tVIRHAFLU) CKiJtDIIUlPi ItLUlK 

P3: 0: phTisal actions] SELECT*. SfDUfl 
[;at4eDt7. CASE 
cliildrfln: Oj DLorphfiu^r YIEiRIHJI 

cb^frg&r>: VDura 

Lt catojory- CliEE 

cata£orjr SQUB 



1: CBtfr^ry- CAEZ 

children^ 0. oLOrptiaiM; K iLIttil FLIU'A 

1: iKirph«n»' RLtr 



SS: projactioD?] fllL 
caTcgoTy: CJiSE 
<:h ildrsi] . pro j act ion? r Efl L 

EDarphau: riRRIHJI 

projection'^ ■ !fIL 
mcixpeiaaa : YIARUUJ 



EDDrpbenia- KAJU3IrL£iFA 
catagorp: VQIQI 

projection?: T 

dAt*; CA5E-PUBKEU: ERGATlVE 
mi^rphBTTiB HLU 



lai 



category^ CA^E 

PS: 0- pliraEal actiucie: SELECT*: ITOTnr 
category i CASE 

cbildreo; Or nrarphfiM" TIIUlIEfJI 
C^tdgOTy : Ef DUX 

1; clUgur^i CASE 



1: catagory CASE 

ctiildren^ 0; fltarpTlfcua^ KAFLDrRfLPl 

category: BDUW 

category: CASE 



3S^ prajftctii^n?^ HTL 
catej^ory: CISE 
chililrflla. prfljBctloQ?: EflL 
mrph^u; TIRRIRJI 
Ut4^Ty^ NDUV 

pre J Act Ian?: KIL 

cal^figory: ERIUH 

pr<]J4ction7r VIL 
sOrphMMi^ KJJttilFLAPl 

catflgorTT IKHIH 

prfljec:tioji?i T 

data: CiSK-KAnKEO: ABSO [itH'l VE 
fflOrphftMt *JiB£* 
catQgor;: CA2£ 

Parsing *((TrHtTNJn CVLJUUlFlU fLLU) (KARDIRRPAl), 

((TIBJlIPJJn (YIRRARU RLUJ (KliRDIRRFA)) S» ungrammit i cal . 
nia prscBdenca BtrUct(ir4 liV luiccnjiactad . 

Parsing *((yrRRIEfJrJ (NTil WVr) ()tAltDrRRPO J . 



Th& precQd^ncfl stmcturs is iii3ci]iiXLec~t«d . 

A.Z.7 Auxiliary Positioning 

Pareiag ([(HAflUJ KAl) (CVl EC!))). 

PS: 0^ phrasal ictions: SELECT*: JiOUI 
c9t4gorf: CA3£ 
childranL 0: DLorph^me: HAFlLU 

catagQrjr CJiSE 

1: leEic;al actions: AUirLIAFL^-3ELECT' AUIILIAAY- DATIVE 

ilU3LILIARY- SELECT i AUULIAFlV-DBJECT 
AyjILIARY-SELECT: AUJi TLIli ftT-3UB JECT 
HI CHT- Al> J ACENT . ( .ITJirLl AaV-SUB JECT 
AUirtliEiY-oaJ&r^T 
AUi;rLIIiFlT-DATTVE> 
data: SYLLiBLES: 1 
■orphfiDMj Ha 
ut4Kcry: JiUXl LI i AY-BASE 

chlldranr U; data: CD ffJUG liTIOT [ & 
narphaiDQ ' YA 
catQ^i^rrr VERB 

1: data: CDH JIK ITI D Ef : 5 

cntagory: TENSE 

»tB£ory; AUXILIARY- BASE 
childrom pro^actj-oci^r VIL 

Gatagary: VEJtB 

chtldrfln: f?ro]ettl&ti'': NIL 

data: T]I^I.-A5S IGKEI? . THEHE 

CAS£-J55ICJ]EI>^ AfiBDLUTIVE 

Gate^ry; CASE 

Ctlldiar^ pTOjACtiDTL?" tilTi 

morphans: HAFlLU 
133 



catagDxy : iflUH 

projection? r T 

dlM; C4SE-«A11*:E[}^ iBSOLUTITE 

category: C4SE 

prcHJflttSonT: T 

category: VERB 

childroA: projection?: T 

dit*; TEJI5E: HOMPiST 

THETA-R0LE5; CTH£[1£) 
TKEME: ABSHLOTITE 

subject: meME 

unrpheBfi . YJl 
cat«gorr ' VEHtB 

pxo} ficticinT : T 
catflEorj^ AUJILIflRY-BflSE 
children: pru] acti i^n? . T 

dJith: TElfSES (unipisti 

ISPECT: rHPEHFECT 

BHUrptlQBL^ . KA 

eataa^ryi fiUlILrAltY-HASE 



(Ul is imgraanL^t i c^I - 

Th4 ALixiliarv has too fev sjll^lilQidn 

PftrflLne KCtNlRLU]) (lU)J USA All)). 

<^Jl] is imgrajutatic al . 

Tn& auxiliary tin> too faif pj-llablfta. 

Pareine (((HAftLUJ (KA LUi) ((YA HI))!. 

PS: 0: phrasal action*: 5ELECT": NOUtf 

[:4t*gcir:r- C*SE 

children r 0; morph^m^: HAHLU 
cats£Dr;r HDTTTi 

i ] Eorph^me ; * AB^4 
CAtfigor^: CABK 



IM 



1: cat€^ory^ AUXILlARt-SJ^^ 

lUlILrAny-DATITE 
AUJ^ILIilRy-EELECT: 

iUKILlAElT-OBJECT 
d4tK: SYLLIBLES; 1 
iurpb4at; A A 
citfifiory^ AUI I LI* BY- BASE 

1: l«icfl] 4Ction«: fllGHT-fl&JACEliT: 

UmiLrAKY-OEJECT 

iUWLlApy-DiTiVEJ 
data SYtLAELES; 1 

m[>rphflnfi : LU 

catifior/. ^UXILIAHY-SUBJECT 

cbUdrfln. Q: data: J^^^NJU^ATICH : £ 
mj^rpti^ow^ TA 
catflg^Tj- VERB 

1[ datfti COtfJUCiTIOIf: 6 

CA^frgtryr TEN2E 

SS: prQjKtl&ci7; MIL 

category J iUlILTAPY-BASE 
childruL^ proJactinnT^ MIL 

d^tir AE113UHEUT- VERS 

cfttagorj: VEftB 

thlldrnn: projectLOD? : ITIL 



data- THETJi'A5SrCKE&. THEHE 

CAEE-ASSIGIFED. ABSOLlTTrVE 
cita^ry: <]1SE 
chtldran. pTOjflCtiOia?: niL 
BLorphene i HAALU 

prcjttttcih?: T 

dfttii CiSE-KAKKEQ: ABSDLUTIVE 

PL^rphuw- ■4B3* 

projection?: T 
categorj: TJ| 

12a 



Ghildrvh: paroj QctiQnT . T 

d4t4: TEN£E: N!]PiPltST 

-raETA-ltOLES- [THEEfEj 
■raEHEr AB5ClLtnTVE 
SUBJECT- THEHE 

toT*B>rj^ VERB 

pro j B c^ 1 cppl' ■ T 
[:atagciry: JlUII LI AJtY-B|l3E 
childralir i^roj&ction? . HIL 

data: PERSDICr 3 

NLJflEEFL- PLURAL 

mtrphama: LU 

catBgory: AU*1 LI XElT-SUB JECT 

pr(j]ectiJ3JiT; T 
data[ TEEfSES; (XClNt^ASTj 
ASPECT- LitfEflFECT 
■orph^mo; ^A 
C&tftjory] iUJLILTflPV-BASE 

PMAing ((CMAaurJ) CIKA LU)) [(¥^ MI)))r 

PS: 0[ phrual tcilfldfl" SELECT*! BflUH 

Chlldr4n- D^ i&DrphenQ. KAJUdlJ 
catflsory: ffOUK 

]: tiorphame. *A£S4 
catagcHrf; CA3S 

l: c*t*E"T- AUIILIIIEIT-&I.3E 

cLildran: 0- Uiical actxcms - J.UK I LI iJty-SELECT : 

|.UTtTLl.lRV-&ATIVE 
lUXILIAIiT-SElLECr: 
AUIILIIEY-QBJECT 
dit4^ SYLLABLES [ 1 
morpkiQaQ. ^A 
cttafiOTj. AUIILIIiftY-&ASE 

1: laJtiCftl actiQua- FLICET-AD3 iCEKT : 

(AyiiLiARv-naJECT 

AUirLIART-D^TIVEJ 
datn- SYLLABLES! 1 



Clto^ory; iUIILIteiY-SUBJECT 

DLorphamfl : Y^ 
Cfttifiorp^ VERS 

1= d^tt; COKJUGATl-orrr $ 

ES: prcj#«]flii?^ EflL 

c at Eigor ) - AUI I LI AElV-S A^E 
children; prdjecticm?: NIL 

data: JlEI^UMENT- VEFtB 

cat*j&rji VERB 

Childrict; prgjftctl.c^n7 r IfIL 

data] THEUL-JlSaLGIfED- T^EHR 

cat4^ry' I^ABE 
cFiildrvn. pr^jQctiDci?^ HIL 
Mrphflcufi: HflfiLU 
c4tB£ory; NQim 

p3^}BctiDri7^ T 

(Lata: CAE B-HAJDCED ] iBSDLlJTIVE 

c-Lta^ciry; CAS& 

pro] act ion? l T 
cat?£Drjr VE3^ 

chlldFEDL JMQJbCtl&C^L T 

data: TEECSE: ICHRPAST 

TM^Ti-nClLES: {THEHEl 
THEME: ABSOLTTTIVE 
31JSJECI" THEKE 
noTptiQiLd^ Yl 
cate^ry: VSlB 

projflctiDnTi T 
category: AUULI AftY-BlSE 
cblldr^rt: pre] action?: fllL 

data: PERSQECr 3 

HUMER- PLLTRAL 



12T 



catBfiorji AinrLIAEll-SirBJECT 

pro j fiction? i T 

data: TEEfSEE^ (naUPlSTl 

ASPFcr- rxPEflFEcr 

I&QTptlfllLfi. pi A 

catflgory: AUIILII-TtY-BlSE 
Paralaft C((K* LU)) ((lURLU)) ((YA ECIJJ) - 

PSi 0: catflgtrT- AUirLH-HY'&iSE 

childront 0: laiical ICtI&Tls: lUIlLIAEiy-EELECTr 

I.U1ILTARY-WTIVE 
AUIILIARV'BELECT: 
AUIILIlItY-P^JECT 
dati- SYLLABLES: 1 
marpbeJud - KA 
catv£9ry= AUIILIAElV-BJ£E 

l: Ifljlcal actlPtlfli EIGHT- flOjACEBT: 

(AUII LI ARY- OBJECT 
AttHLXiHY-DlTlVE) 
iAtti SYLLABLIS^ 1 
norphvmb - LU 
cat&aory: ilUlILlAElY-SLFBJECT 

1: phraBll tctiOM: SELB^T*: SDlJlf 
C^tegaxjr <;ii£E 

childranT 0- morphflmer [IBFlLII 
cat*fti?ry: BOUH 

1^ mflTptiflne. ^AB3* 
catflfiOry" CiSE 

chil-iran; 0; data: CETFTJUCATTIDW : B 
niorphflmn - YA 
ta^-egory; VERB 

Ic data: CQBJUCATIOB ^ B 
iDDrphsn^ . HI 
categflt/: TENSE 

Sg: prnjfiCtlcn?^ EmL 



categchrrr AUULTIiftT-BJiEE 
childraiiL pro j act ion? ; HIL 

data: AElCU}^£fn~: VERS 

cftttgtry^ VEHB 

children L prajaction^i ML 

dttti. THETA-ASSI^flED: TKCTtE 

CiSE-USIGNES^ lESQLUTIVE 
CbtQgc^^J^ CliSE 
chi LdraD l pro j ftc t loe^ i !fIL 
iih9rp!i€m«. KAJILU 
cateKOjy: VQUK 

projoctioD?. T 

IDDTptlSKfi I "iBS* 

careg^iry : CASE 

cattfcary^ VERS 
children: pmjaction'': T 

dbtA^ TZKSEr EftlliFiST 

THETi-TtOlES: (THEHE) 
TBFHE" AHSOLirrrVE 
SUBJECT, THEHE 
DarptiBiLd ' Vj^ 

prn] action?: T 
categorr= AUULI AEIY-BISE 

»ti. PEASQPI. J 

jnJMBER: PLURAL 
Pfirp&fc^fl ■ Lit 
catBflQry: AUII LI iAY-EUB JECT 

proJactionTi T 
data. TENSES. (ROBPAST) 
A5PECT: IKPEJttECT 
bOrphwuB -r }ik 
CAtflftorr? AmiLIARV-BASE 

Paxflina *<(OURLUJ) ((Yi MIJJ C(XA LUJ)). 

CCIHABLU>> ((YA KI)J ((EA LU>H Ib angraaMticBl. 
nia aoiiliazj is not in tti* propar poflltlont 



]?9 



A,2,8 Free Plirase OrJ#r 

Plrtlng (((EfGAJULU RLU) U* ^^ tlLiJ) C(PUlfTA RKDJ CCKUfLDU KUl) 

FS: 0; ^rftflal actioDB[ Fi^rj-CTt : hqIJA 

CbtGgQry[ HDUH 

1: ■orphdEoa: HLU 
Cflt&gory: CASE 

it category: ALFIILI ABY- StfB JECT 

cLildrea: 01 L«lc*l actions : AUII U * JtY-S EI^CT : 

jLUSILrAHT-DB^lECT 
dat^^ SYLLlBLESr 1 

ciT^ajorj] iUJLILrAJlV-flA5E 

li data^ ST1-MBLE3- 1 
■DrphQQie . FiNA 
caMjcry: AUII L I ART-SUB J E^TT 

5r liatai SYLUBLES: 1 
morphaBo: EILA 
catflg^rj. JLUXlLI*JtY-[>ATIVE 

2: ci&tajoTjr VERB 

chtldran. \}-, d*t^: CaWJUGJiTrQIf] a 
norpheiDe: PUVTA 
category: "JEflE 

It ^a^M CaVJUGATIQH: 2 

CJiLftgOTy: TEMSE 

&; phrM»l actLona: SELECT'; ACMJU 
cate^rj! CA^E 
ctildrfin: 0; [ncHrpJuims ^ KUFlDTJ 
cat&goryr BQUH 

1 J iKrpbiQQLQ . KU 
Cftttaciry- CASE 

4: phrasal actictis: SELECT*: HDUH 

m 



c:aU*g*TT: CA3E 

1: ncrphfiom: *JiBS4 
cattiJoryT CASE 

55; proJBCtlDh'^ BIL 

cbllirin; pro j action?: MIL 

dhtL^ AfLClfflEUT- VERB 

utBgOl-yJ VE91B 

tklldron^ projoc-Lion?L BIL 

d»t*" TEETA-ASSIflHEO: XGEHT 

CASE-JiSSIiiHEU: EHVjATIVE 

Ctt«eciTy: ClSE 

chiLdruir |proJ*ctton7r HL 

datn P£ft$nH: I 

HlfflBER; (SIHCULIlRJ 

aiorpb^Die: liGA^FULU 
citflg^ry NOUK 

frD]flctiorLT: T 

jm: CASE-TIARKED. ERGATITfi 

^Ategorj; CJlSE 

children: projflctSoftV^ MIL 

dlti: THETA- ASSIGNED; THEME 

CAS&-A55IGIIED: ABSflLUTIVE 
c;Lt4£ery: CASE 
cttilireD. pro J act Ian': MIL 

catfigQryr flQlTN 

proJ^CtionT; T 
data] CASE-HARKED : 

IBSDLUTIVE 
ijisirp-hoTDa : *ABS* 
C^to^ry^ CASE 

pra]*ctionT; MIL 

data: Trt&TA-LINHEB: PATH 



IJI 



cblldrfrn^ pre ] act Jen? ^ HIL 
imrpbems: KUEDH 
cbtfr^ary^ SQLIH 

pro] Action"^' T 
data: CiSE- ASSIGNED : 
DilTITE 

THETA-A3SIGNZ0; 
PATH 

catagorf- CIlSR 

pro J acTioFi'' ■ T 

THEXi-RQLES r tiaOlT TI4£HE 

FJ.THI 
lCE3rr: EFLGXTIVE 

PITH; DATITE 
SUBJECT" J.3EST 
flflJECT: PATH 

nerpttAiM ' PIDITA 
c-ateEcry: VE3tB 



projectioD?: T 
Citagc-rp^ AtniUAFLT-a*SE 
chUdrui: projection? : NIL 

narpboiDH r ELi 

Category ^ A^ICILIARY-DiTIVE 

prii]Bction?. NIL 
d&ta. PEE^OX. 1 

KUHBEfl: SIHGULAR 
pu?rp>niliifr: MIA 
catsgory: A UIILI AEir-Stm JECT 

pri>]flctiQD?: T 
data. TENSES: (HOBPAST) 
ASPECT- IHPEStTECT 
mc^rphflis? r KA 
catefiorj: AIHtl LI iEY-B ASE 



132 



C^pukt;, BHiH). 

PSr 0; pr.rBE.1 ^ctlon»^ SELECT-: HOUK 
cacseory ChS^ 
chlldcan: 0. mirpbamO^ XQ^JULO 

l: BOCpllBIBB: HLU 

liStoBoryi ciSE 
I- ^7,t.aon JLyjULIAHT-SUEJECT 

t«.S°ry= lUJlLlARl'-SOElECT 
7. data: SYlUflLES: 1 

2, phra-il a.^L^na: SELJX:T*= HDUB 

catflgcHrjL CASE 

a- phiM^l actions: SELECT.; HOim 

catatcry; NDUH 

^IdFEEL; 0; data. CDHJtKiTlClHL 2 
morptiamfc: PUNTA 



133 



l\ data- CDIf JUGATECB ■ 2 
■orptieoiQ RNl 

cataj^^Tj: JlUII LI iRY-B ASE 
Chtldrani projactlon?" KIL 

catflgarj: VERB 

Cftilirwit proJactiQB?t MIL 

dfttfl. TH&Tfi-A^aitlEJHl- ICEifT 

CiSE-iSSIGHED: EflClTIVE 

childr«n: pnn j dcticin? . NIL 

KlFHBEn^ fSltrCULAFL) 
mi^rpharnQ : I^GIirfUlU 
catfigor;: NOTTir 

^Atg cltSE-HAaHEQi EFLGATire 
morpbama- KLU 
C4r«£ary^ CASE 

iiata^ry : VERB 

dAtar TKETi-A^SICinED: THEHE 

CAS&-i^&5IGNK[i; AfiSULUIIVE 
CAtft^ry: CAEE 
ctu-ldrBa; pn? J €^ 1 1 ?ti7 : NIL 
Fqj^ri^cm&. KAItLI 
category: SOUJf 

prcijectioi]?: T 
data. CliSE-HliRKEP; 
iBSaLUTTVE 

rwrphiMi&r ^AfliS-* 
categciryr CIlSE 

pro J€ctii?nT : ECIL 

dAta^ TKETi-LIVXED] PiTB 

cht4£ory; CASE 



l»i 



pro j action?: T 
data: CJi^E-ASSIGHED: 
DATIVK 
CA3E-HAFLKE&T 

DATIVE 
TblETA-J-SSICnED 
P^ITH 

cato^ry^ CASE 

preii#ctiDEi7r T 
data: TZJaSE- NOtfPliSr 

TKETA-aDLES; (ACtENT THEJtE 

PATH) 
AGEin-[ EHQAnVE 
TIEKE: ABSOLtnrVE 
PJTH: DATIVE 
SUBJECT- AGEm" 
OBJECT^ PATH 
IKTph^me^ RJlfTI. 

prQ]flctiDQ?: T 
tA^aSoryi lUJLILlAElT'EAaE 

childi^fll pio J Bcti &n7 ^ NIL 
EhDrpbisDLa . ^Uh 
titigoTy: iUIILrAftV-DATITE 

projection?^ BIL 
dbtl- PERSQEf: 1 

IfUffBER- SIffGULUl 
■iorph«mfl - Itif 1 
CAtagory: AUILLl Afil-SUB JECT 

pTojactinnT T 
dtct: TEKSES^ JSUHP^ST) 
ASPECT" IMPERFECT 
narphsBv: Kk 
^ataeoTjr AUHI U J. A'f- aiEE 

Puiinft ((CKUiUS" *^> (KA HVil ILL1)J (CPUirTA MI]J [[PfQUULU TtLXJ)J 

135 



FS: 0^ phrBsol &{]txQiiB: SKLECT+: ETCUli 
catBfiSty: CASE 

catQgQir: CJlSE 

L, C*t*gi5ryi iU^lLIJ-RY-EUBJECT 

chtldran: 0: lexical iCll[>tar I.U11 LI iny-SELECT : 

AUXILIAIlT-dBJECT 
data, syllable:^. 1 
borph^rfiB : HA 
caiagory: AUirLTAnT-6lL3£ 

Ii dii»; SYLUBLEiS: 1 

c4te£oty: AUIlLlAPY-St^BJEc^r 

Si dat^: SYLLABLES- 1 

parpboDft! itLJi 

catafiorj: iUlrUARy 'DATIVE 

^t c At •gory: Vem 

children; 0, d«a: COHJUaATIDH- 5 
■arpheiDe] FUKTjI 
catflfor^: VEAB 

1: data: CD Ef JUCATIQH ; 2 
■grphetP^: RHI 
cfttagory; TEIC3E 

3: ptiTcsal acticiDK. SELECT*: ItOUH 
category: CASE 

Chndr^n: □■ KFptitiiu: NCAJULU 
M^tftory: MDuy 

1: porpbiana' RLU 
categ^r^i CASE 

4: pbtaaol actlcnh*; SELECT-: flDUN 
category 1 CASE 
children: Ol cfr^irpr.ena: KIRLI 



13^ 



It ftCirpIUMii*- ^ABS» 
cat^gcir?. CASE 

S3: pr^jectiDD?: HIL 

CBie^Tf JL[r£ILIiRY-BiSZ 
Gbxldxen: projftctioD?: NIL 

categajjr VERB 

chiLdriEL- pro j fiction^' IfIL 

CiSE-lSSIJjTlED: EJU>ATIVE 

children L projection? [ KIL 

4»»; PEa^PN 1 

ECUMEERr (SIBCULIJl) 

catafiory. ifOUH 

projection'': T 

data: CA^E-HAKKED ^ EAGATIVE 

Hrpb4E44^ HLU 

cata^ry^ QASt 

proJ4ctioa?[ T 
CAtQ£orrr VERB 
cblHrwi^ prfljaction': IfIL 

data: TBETA-ASSIGHEB: TBEHE 

CJiSE-iS^JQlO. ABSQLliTlVE 
catfigdiT- CASE 
ctiildranr projocticih?: NIL 

UtorpheiLfi: XARLI 
catBgQjjr KQUPl 

projection?: T 
dataj CISE-HIMED; 

AEEQLUTTVE 
HorpheiDer *AB3* 
categQrT' CASE 

p^jectioD?: NIL 

datil TBETA-UlREO: PATH 

tatagciryL CASE 

childrsD. projecticn?: HTL 



13T 



hprph^mo: KURDU 
cata^ry^ HOUW 

DATIVE 
THBTjl-JLSSrGffEU; 

PJlTH 

THETl-ROLES : CiGEHT ITaEHE 
PiTH> 

iQ&tfr. EEUjinvE 

nEME JtB^Qi.tn'ivE 

Filfl: DATITfi 
SUBJECT" AflEBT 
OBJECT. PATH 

imrptitM^ PliHIJi 

cata^ryi YEKB 

pToJactton': T 
catQgHpry. JilH^I LIA AY-BASE 

chlldrfiDi pTo] action? : IfIL 
■crptLfffl* : FiLA 
catogorj- lUXILIAny-DATIVE 

projection?* HIL 

data: PEFLSOH] 1 

KUfEBER] SlHUtrLAR 

catagoryE A UULIJlRT- SUBJECT 

pxaj&ction?: T 
dita- TEB5E3" (NOKPAST) 
ASPECT: IHFEfiFKT 

morpbiazL^ : Kl 

CBtejory AUJtl LriRY-H 13E 

Paraing (((EARLI) (U Ml RLO) UBGAJULtJ RLU> | {fKURL^U KU) ] 



ns 



P3i Ol phraaal actLonB. SELECT+' MUM 

catBgorf: VDUH 

CAtflgory: CASE 

It Ct.t4i^Ty-. AUIL LI ART-SUB JECT 

cELildren: Ot laaic:*! tcili^lli^ iUTT LI I-JIY-SELECT [ 

iTjiiLLjiiiy-oajECT 

dit4i 3VLL*flLE: 1 

hfirpkifiBa' Kl 

catBgc^ry. AmiLI*HV-BASE 

1: daM[ SYLLABLES; 1 
ncr^4Dt: ^l 
category: A UTL ILI ART- SUBJECT 

2: datai SYLLABLES. 1 

tattt^rj] JUI I LlApy 'NATIVE 

iJ^ phraE&l a<]txanB[ SELECT*: K^JH 
C4.i:*£4ry: CASE 
chlldxan: 0[ iDDiFheiDB: SQAJULU 

1 : DLorpbemB r EILU 
Cktqg&rj'^ CISE 

S] pbraBal actiooB: SELECT*: NQLTH 
C4^4^ry' CASE 
children. 0: mffrpll*B*' KlIRDU 

1: DLOrphOEDB: KU 

catBgDry[ CASE 

4i catagorrr VERS 

children ^ d^ data: COITJUGlLTIOir; 2 
DLorpheme: FUirU 
catv^ory; VEfl.B 

1, ilal4^ CQV JUCA-TIOfl I ? 
13!} 



Cttagprj: TENSE 

5S: prDJ€cticin^: ECIL 

catBflory: jmirLTAny-EASE 
chlldraci; pr<]j4Cti4ti?: NIL 

data: ARGUHEFTT: ITERS 

tategorj: VEftB 

ctiildrsn: projection^: NIL 

d*ta; THETfl-ilE5IGll£C: AetEBT 
_r CASE-lSSHiWED; ERGATIVE 

■icatagoTy: CiSE 
ctiildr*}?! projdctLc^nr; itL 
d&ta[ PEIl^Cn- 1 

NUH&ER- CSTteUUM 
norphflmo] BGAJULU 

piojflc^l&n^ T 

data: CA3E.-HAFLEEII: EFlGATIVE 

Il0rph4b4; ALU 

cate^r?] CjISE 

proJBction?. T 

chlldrflt]. projBctioii7r BfL 

dB^A: THETA-IlSEIG?^!^?^ THEH& 

CASE- ASSIGNED; ilBBQLUTIVE 
c^tofory] CiSE 
chll4T«n^ prthjfcctititi?^ HIL 
b?rph«tiLe; KARLI 
C4te^?y- UOUV 

dfttar CASE-WABKED: 
AflSOLl/rrVE 
miirphftmfii +ftBS4< 
catdfitjiy^ CASE 

prcijacT:lon?T IlL 
data: THETl -LI EfKlB : PATH 
category- CASE 
childreii: projection?: NIL 
mchrphAffiG : KUFLfMI 
categc^^y^ BOUTi 



i4{] 



DATIVE 
Cil5E-H*BXED: 

DATIVE 
thetjl-i,eeichEd: 
PATJt 

CAta^orjE CASE 

data; TEfGE. ETDNPflST 

THETi-MLEEL [WEMT ™eHE 
Pl.TH> 

A^EHT: EBCI.TIVE 

THEME- IBSDLUTIVE 

PATH. LUTIVE 

SUBJECT^ AGEJIT 

ODJKT E^lTt 
nhi^rphfina : PUKTi 
ctttfgcry^ VERB 



pn3JBCtlon7' T 

Cht4g«rf: AUIILIAnY-EASE 

chlldrftTi? prfljftctlcpii'^: BIL 

ifl£hrphflmB : ELA 

Ctttg&rr^ AiraiLIXRY-DATire 

data^ PERSDH: ] 

)fL7[BEE: SllTGULAFL 
K^rphema : lUf A 
category: jIUJ: ILI AflT-SlTB JECT 

projitticn?: T 
data: TEfCSES; (HDNPAST) 
ASPECT- IMPERTECT 
DorphfiDie 1 ITA 
Cft^t^ry: -AUXiLiAny-BA^B 



A-2.9 Ai^ffiuncdt Ideuliflcatlnn 

Parai&a ailKlJULU ElLtJ} (Kl MA ALjI)) [(PtlHTA HU1» KHUHDU KU)> 
(IKARLIHI. 



141 



PATHl pMjflcti^nT. NIL 
jmrpbame : KURDU 

THEEE: projflcti^ft"?- 1^ 
EdChrphom^: EARLI 

AGEBIi projsctii^n?: KIL 
data: PEtlSDN. 1 

NUMBER : (SIHCUUIO 
Br?Tpb°« H13AJULU 
cateftisry. HDlffl 

Paniiji£ (((HGJlJULU ElUJ> (KA AAA KLA>J UKULIDU KU}> ((KIMIJ) 
tCPUHIi Hltl)»- 

Pjmft pr&JictlMT- EflL 
DLorphaow] EUm^U 
cttfrgoryi NOW 

TMEHEr pru] qCtl^n"^ ■ tfTIi 
TfifbTpTiftTfift ■ KARLI 

cAtagciry: HDUN 

AGElfT: pro3BCtioti?r HL 
data- PERSDH] 1 

mi^rphem^; NGIi-IULU 
cat •gory: MDLJff 

ParBing (({KURDU KU] tKA BMA BLA)) aPUNTA RHI)) ((BGAJULU RLU)) 
{(KARLI) I). 

PATH: proj^cti^nT. NIL 
iTwrphwflfl: KURDU 
catagcfry. KQUN 

TEIEME: pTOJscti.on'^ ^ NIL 
Kirptiama: KAJtLI 
Cit*g*rj^ BOlffl 

ACENT: projection?; liTL 
da^aj PEASaEf: 1 

N1JH15ER- (^INCUUn) 



Iti 



ttPUTfTA RK[)>>- 
PATH^ pr^Jictlon^L IfIL 

cat&jory] NOUff 
THEHE: pro J acti cni7 ; HIL 

c^itflEflfy' HDUH 
JlCEHT: projection?: KL 

EfUHBEEl: CSIEJCULAFl) 

m&rphgma: WGiJULU 
catagory: HQUV 

A. 2.10 Null AiixIfUry Camponentfl 

PS^ Q= phrasal actions: SEL£CT+r IfDUH 
category; CilSE 
cbildreDi 0; nnnrpheioB. TtAniU 
^atagary: iraUV 

1 : DLorphuu : ALU 

1] iQiical actions: AUJLTLl AEIY- 3 ELECT r AUlTUART-DlTrVE 

AtlllLlAnY-SELECT: AUlILIAElT- OBJECT 
lUIILlABY -SELECT: AUl I L T AR/-^ BLm JECT 
BTGirr- ADJACENT: C IlUII LIARY-SUB J KT 
AUlILTARy-OaJECT 
lUIILIAIlY-IIATlVti 
data: SYLLABLES: 1 
j&orpbeiaaL KA 
citsftpr;: ilUTilLIAFtr-BASE 

2- catB^ry: VEFLfl 

HLhxldran: Ol data: C^^NJIUATIDNi 3 
^orphema' NVA 
catsgc^FT. VER£ 

1« 



1= dfttk; CONnHJAtHJH: 3 
narphoms i HYI 

3l phjual a^tioiiA: SELECT*. VOUV 
children: 0: aorphsne^ KUnoU 

C4it4iKory^ Cl^E 

ESr pro j edition? ^ fliL 

catajory: XUITLI iAY-BiSE 
<]hj.ldren' projactLocTr HIL 

datAi ai^(;h^t^ verb 

catagory VERB 

childr€D. pr? j Bct IchnT r EfIL 

CASE-ASaiGWED: EEICATIVE 
cata£-?ry. ^A^E 
childTen] pTD]actxQn?: IJIL 

Di£>rfhanfi = KiALU 

catQgQry[ IDU}] 

data: C A2E- PUnKEI^ ; &HCATIVE 

proj4Ctlcn1'i T 
category: VEflB 
Children: prcjdcticn?: V7L 

da1;aj THETA- JSSI G MEH ^ THEME 

CA5E-lL5SrG2fEQr Afl30Lin"IV£ 

childrBHC projecti^^n?; NIL 
Harph&iriQ- KUR&U 
cittdgory: SOLJU 

prfl j ett i<3 fi"^ : T 
datBT <;iiSE-MliF^ED; 
iBSOLUTIVE 

catQ£?ry: CAS£ 
144 



dattL TENSEt MOSfPiST 

TKETA-BOLE&i (A&ENT TEIEME) 
AGENT- EJmiTrVE 
THEHE. iSSDLUTIVH 
SUBJECT: A&EMT 
OBJECT^ TitEPE 

ca^a^^r: VEEIB 

projection*' T 
catejoxy: AIHCI LI AUT-BASE 
childran. projection?' T 

data: TEKSES. (MOBPilST) 
A3Fb::t. inp£HfECT 

BHurptiQiLe : Kl 

t»t*BO ry I.US1 LI Aity-B A5E 

PuBiufi fftiGiJULU RLU KA MA)) (CHYflllYD) ((KUJUSUJ)). 

PS: 0= phrasal atttohri; SELECT*': VDUH 
i]&tft^ry: CiSE 

children: 0^ ADTphanci- N^AJULU 
E;atfiga]:T: UDUH 

cat«£orr= CASE 

1: »tQ£ory; AUILI L I AE17- BASE 

cUJdren: Q- lejical actloiLB: LUXI LI UT-SELErT: 

AlHtTr-TAJLY-DATire 
AUiriliKY- SELECT: 
lUXlLI in Y- DEJECT 
datai SYLLABLES: 1 
uorpheina^ KA 
cats^cry: AVlIIiI J.RY- Bil&E 

1] iHiical artioi]B7 RICFlT-AriJACEffT: 

UUlILIAflT-DBJECT 
AUXILlAfLT-EATCVE) 
detA: 3YLL*I!LE£: 1 

CAta^ory^ AUXILIAFLY-SUBJKCT 

2. qatflgqHry: VEEI& 

IhU 



claildr*!!' fl: data. CDB-TtKiTini]: 2 
IDDTp1l6>fl r HtA 
citftgor?^ VERB 

1: data: CO N JUGITI D ff [ 3 

citBgory; TEUSE 

3t phTMal attlona- SELECT+r HOUP 
catBfttry: CASE 

children: Or iwrpKaMi imoU 
c^t^^ry: NOUN 

CKtegChTjL C45E 

SS: projactUnTi HIL 

catBftory- lUIlLriBY-BASE; 
Chlliiaa: preJ»ctL&n^: ITIL 

dEtai iRaUHEBT- VERB 

»I4£CprF^ VERB 

childroti: prflifiction7; MIL 

diita. THCTjI-ASSICHED: AGEEfT 

dASE-BSSicif&ti : ERcanre 

CBta^ry. ^A^B 

Childran- piojactli^ti"^' NIL 

^4^4.: PEFL£{]|]: 1 

HUHBERl (ElBGULAHl 

iDQryilisu: EJGAJULU 

-latQgcirTr VOUH 

project icn^: T 

data: C JiSE- HA R«Kt ^ ERCATI7E 

-csate^ry: QA^E 

catflfii^rj r VERB 

cblldr«Ei: projection?: MIL 

ittaJ THETA-A5SIGNM: THEME 

C:iiS£-ikSSICEfEI]r ^BSQLUTIVE 
categQrr: C^E 
tbxldxBQr prajactioD?: BIL 
morphBTTiB- KUflQU 



1« 



iBSflLtJTIVE 
[^atagctry^ CASE 

dAta: TEHBE; HOtfFlST 

TffETA-atlLESi (AQEITT THESE) 
tCEETT- ERG ATI VE 
IHEJE: AESnLOTiVE 
BUBJECTI: A<1EHT 
OBJECT: THEME 

c^tigory: VERB 

Mt*g&rr= HiniLlftHY-BASE 
children^ proJeCi^ion^ : Tl^L 
dbt^i PEH^OH^ 1 

UUMBKlt SIR^jULAa 

<:ai.*Sory- 4U1ILI AftY-SUB J ECT 

pro]*cti<i ti"? ■ T 
datai TMSEE. CNOMPASTt 
ASPECT- IMPERFECT 

tatoBflry: AinllLIAftY-6*SE 

PS= 0- phr*Bal iCT^Iodb: SELECT*^ NDUH 
childranr Q' Borph*™. NCIJULU 

c&tegnrT' CASE 

1- i;ttaMrV^ AUI I UAilT-3tfB JECT 

thlLdrs^: Oi lexical ^ctionfl^ AUTILl ART-SELECTi 

AUlILIAflV -DATIVE 

iat*^ BVLLABLlTi" 1 
borphQDiB - KA 

H7 



QLorpheioQ r HITjI 

2t leiical actional FLlCHT-iD JACEKT : 

data. ^VLLAPLfS: 1 

BLorf-beme : HGECU 

tutitgory: lUXIL lARY-Ofl JECT 



2: CAtti^rj] VEEIB 

nwrptiaDLe . ETf A 
CbtdgOTy VERB 

1: data: COHJUCATI OH : 3 
saTpkiBma : ETTI 
Cbt4e4ry. T£VS£ 

3^ phTftBal «[^tE[?nH: ^ELfrCT*: HUUH 

thlldron. d toi>rphiWfln- IfVimtULU 
category- VDUn 

1- morpheiDBr *ABS* 
catQgQxr: CASE 

SS: prajfiction?' NIL 

catBgory: AUJ ILI ART-EliSI 
childrfrTi: proj*ctScin?i HIT. 

data: AFlGUHSNT: VU^ 

catBgurjr VERB 



data- TiJETi-ASSiaHED- i{]ENT 

CA5E-AS5raiiE&; EftQATIVE 
Cfltftgary- CA9R 
chlldrtn- pr^Jscti^art?; UIL 
data. PEBSCTIT; 1 

)fUHBEEl: CSIVatJLAEl] 
mchrphenB! KGAJULU 
catBgar^r ITDUH 

14S 



c-atflgDrTr CASE 

children^ pnHJectioci?, HIL 

dataL THETA-JiSETCirEDT THEHE 

CA5&-A&SrCI]ED- ABSOLtJTITE 
□at&^rj: CASE 
chJr^dTdn^ proJectionTL HIL 
data: PEnSQT]^ 2 

fltJMBEB- (SrHCULiRj 
moTphaM. EC7UKTULU 
^atd^ry: ND'J}] 

pro j act ion? : T 
dtta^ C*3E-MUtXED: 
IBSQLUTIVE 

mtHrphecna : +AK+ 

C4tflgory: CA5E 

proJncil&nTi T 

dita: TEEfSE- EHIMPAST 

■n^ETi-nQLES. fAQEEfT THEHEl 
AGEiCT^ EMATIVE 
TBEHEr IflSOLUnvE 
SUBJECT: AGENT 
OBJECT. THEHE 

iborptiVDWi Nil 

Gata^ryi VERB 

ptoJactionT] T 
cataectr^; AUI ILI Af^T- BA^E 
cbildr^n! pre j action^: IfIL 
diT;4 PEF^QN- 7 

NimBER. sirJi^uun 

catBfiorrr ALFTI Llli RY ■ D B-TB^T 

proJflctlonT: HIL 
dati; PERSDH: 1 

HUKBEH: SINGULAFL 

IKrpti4tD4: f^A 



d4tir TENSES i (tfnNPAST) 

ASPECT^ rHPERFECn' 
norpbeme: KA 
CAt&go^y- AUirLlAFLT-EJiSE 

PftTBing •{{CJfGUULy HLU U BGXU)) ((WYA EfTI)) C (NYUHTULU) ) ) . 

{((IKiJULU niU KA BGKUH ((IIVA ItVLl) ( (SYUimJLU) J > is ujigrMmiatLcal . 
The DEFAULT- AUXI LI ART clitic dae& not agrse In p4»i^n wUtl th* 
SUBJECT. 

Parains * ( ( (MGA JlTLU KLU KA JtlfJ.)> CtUYA HYI)) ( (KVmiTULU J ) ) . 

KfHGAJULU RLU Kl. MA)) KtJYA HYI) ) (( ETTJUTirLU) ) ) U trngrawns. 1 1 cil , 
TlLfl DEFlULT-iUIILIAftY clitlC dOfS MT afira* in parson with the 

A^Zai Null Anjiphora 

Pwsin^ (CtlTfA KYI KA) ) ([KUilDUt)>, 

PS. Pr <«*gMyi VEKE 

childreDi Or data^ CONJUC^AriD)!^ 3 
jtftfjitidpi*' ffYA 
ciT«i£cry: VERB 

]- datA- CDVJUCATIOHi 3 
K^rpbiQQie; Tffl 
<:ate^ryi TENSE 

1: lerical actions. Aim LI ARV-3ELECTi AUII LI ARY- DATIVE 

AUllLIARY-SELECT; AUX 1 L3 li RV- Q & JECT 
AlfllirUY-SELECTL AUIl LI ART-SUBJECT 
niGHT-AUJACTEflT: {ArJlILIAFLV-SU&JECT 
AUirLIART-DBJECT 
AUIXLI ART-DATIVE] 
d*^ti= SYLLABLES: 1 
■orphana] Kl 
C4tif£cry- j,u?LiLiAny-BA9E 

2- phrasal actioEiB; SELECT*: WDUTJ 
i:ai:4g£iry' Case 

chLldr4t], 17: rtdrphamfl- KUflDU 
catBgar^r ETDUH 

150 



1; porpheiDBi *iBS* 

SS: pi-cijection?: MIL 

cFtSLctren: pr* J ecti on? HIL 

data: AlL&lfflEHT: VERB 

children: pr^Jftction?^ T 

flctioJifl: CASE-A^SrCH: EPC4TIVE 
THETA-Jl55IGNl AGEBT 

childracL- projections r HIL 



dttn THETA-ASSIGflED: THEflE 

CIS E- ASS I CITED: li^DLimVE 

CAte^ry; CASE. 

QhildrenL pnsjftctloiiTr HIL 

Cfl-tG^ory: BDUB 

prfljttctloh?^ T 

data; C^JlSE-HJlFLHKD: 
A^QLUTIVE 

min rph^ne : *ABS* 
cate^orj: CASE 

projoction7= r 
data: TENSE. taQHPiST 

THETl-RDLES: UCEBT THHE) 

iCEffT; EMlTIVt 

THFM: ABSOLtJIIYZ 

SUBJECT: ACEJIT 

OBJECT" TMEHE 
BorphflBifl' HTA 
category: VFflB 



proi&cttOft?^ T 
oatBg&rr[ AUIlLIifiT-Blse 
^hiLcir&a: projection*: T 

datt^ TEHSES: (tfOHPAST) 

thtasDJ?. AUlILliRT-flASE 



151 



chUdrac] C: dataj <:0N JU^ATI DH : 3 
cate^r; : VEAB 

1[ data: CaHJTTGATIl]}] : 3 

cata^Di;: TENSE 

Ij laKical actionEir AUITLI ABT- SELECT [ AUIILTAEIT-DATIVE 

AUllLIJiFLi-SELtCT- XUl I LI AJlY-flB J Ei:T 

auiiluleit-select^ IiTjjli LI art-sub J EO" 

niCHT"ADJ*CENT- ( iUJI LI iRY-5UB JECT 
AUlILLARV-DBJKrr 
lUlILrARY- DATIVE) 
dAli; SVLLIlBLES: 1. 
■orptL^DLa : U 
cfl.i*ftOTy: AtfX:LIUtY-B13E 

SB: proJ«ilM?- KIL 

category: IUXILIAFlY-BA^E 

chlldrAA: proJflC^l^ti?; NIL 

data- AElCUHEirr- VERB 
catagory: VZflB 
chlldTvi^ prflj*[itlcp?: T 

actions: CA3E-AS£rGB] ERCATIVE 

rHETfl-li^SICN^ ASENI 
catfljQijr VERB 
childrtut projActloh?: T 

Bctlonfl- CA£E-J£SICir^ ABSnLliTIVE 

TRETA-ASSICfffr THEffE 
dAts.: TZtfSE- NQNPAST 

THETA-HQLES: (AGEEfT TT^EHE) 
iCEHT- RRGATIVE 
THEF!E^ ABSDLTJTIVE 
SUBJECT: AOENT 

iDDrpliBBa: ITYA 
cqLdgory VERB 

prQjectiodT: T 
cat€gDT7[ AiniLIART-fllSE 
chlldrBUr projecti.oaT'r T 

data: lEBS-ES: lUOHPAST) 
ISPECTl IKPEKFECT 



■wrph?b4i HA 

cftt^^ryj iUIIUAIlY-BASE 

P^rrtlnfi CUHA HNA KCKUJ) UKTl ICyrS) (CSYUlTTULXJl)}. 

P3: Q" CRttgCirj: lUJLI LI ARV -SUBJECT 

■ihildran: 0. Ualcil ^C^Urtfl^ IMI LI AhY-SELECT : 

llJIILIAnr-UATtVE 
d»TA: 5YLLABU3 ] 
porptLBOiaj EA 
caT^ft^ory; AUJILIABy-EiSE 

1: data^ SYLLABLE. 1 
■grf^rL4Di4: UIA 
cate^ry: AUU LI AFlT-SUB JECT 

(AUJLIUABT-DiTIVE) 
data: SYLLiBLCSL 1 

^atdgor/- AUXILIAIty-aDJECT 

childrasL i^^ daUL CDNJDGlTIQTi: 3 
DPrph«mF : VVA 
cit4£ory: VEfU 

I: il«^»^ COKJUaATIQI: 3 

iwrpbHue . PTf I 

2. phrasal actLi^nE: SELECT*: lOm 
catagiDry- CISE 

children: 0: niarpheiDa. ECYUflTULU 
C At agar; ^ WQ™ 

Ij ■erpb^mo] 4JiBS=* 
catflgory: CASE 

£S; prPJ4ctic?n7: TilL 

catttjory: AtUI LI AKY-EASE 
chj-ldran: protection? : VIL 

dU4; AnCrETHEirT: VEFLB 

categciry: TERB 

Cbliar*ft: prQjtci1fln?j T 



1S3 



TBETIi-iia^rGB: A5EHT 

chlldrfln, pvij action?: RIL 

data. TKETil-JiS3ItTrfE0 THEflf 

CA3E-ASSrGHE[>r ABSOLUTtVE 
category: CASE 
cbildran: proj«ctI?i]?l HIL 
dAta: PEE^OH: 2 

KUKBEH; ^SIHSliUFL) 
Hiorpham&: FflfflTULU 
category. m\JU 

proJactianT: T 
dfti^a: CASZ-MAFLKED : 
ABSOLtTirVE 
iEh:irpbi3a« : *I.BS4 
categorj : CASE 

data: TEHEE 1 NQHPIi^T 

THETA-EDLES: (AGENT TTIEHEi 
fl[j£NT^ ERGATIVE 
THEKEi iBanLlITIVE 
SUBJECT: AGENT 
OBJECT: THEME 

iK^rphama] ITfA 

c^at4|cry: V£n& 

prnlflCtlfllT^i T 
cat#Bory: AUIILIAftT-EJiSE 
childxaD: pn^je-^tlofi?: MIL 
d&ta- PERSOH^ 2 

CAl4£0ry^ JlUILLlAFLf-a EJECT 

prciJ*ttitnL?- HTL 
data; PEE^GH: 1 

HUKBEH: SIHGULAE 

category: JIUITLI ART-SUBJECT 

prtj] action?; T 

datar TEN5E5: (NOHPABT) 



164 



ASPECT^ IHPEFLFECT 

cat^g^^ry. liUllLlAJiy-fiASE 

ParfltDfl (((HGAJULU RLU KA IWA WGKUJ ) ((HYA Em))J . 

PS: 0? phrtB^I actional SELECT*: SfQUl 
cit&goryj CiSE 

children" 13: raarphoiDQ' Efi^AJULU 
catsgQrrr VDUN 

]! norpheoia^ ALU 
catfJgory: CASE 

l: Ct^4£C^ry: AUTILIiHT-SUBJECT 

childrflD? 0; lixlcfil fCtiosB- ATIIILIAAY-SELECI: 

AtlllLIlAY-DiTIVE 

data: SYLLABLES^ 1 

AQrpbflmH -r KA 

CAtfrjory; AUirLIART-BliSE 

DDrptiaB« . FLNA 

cBtflgDry: AU:tI LI IRY-SUB JECT 

2: l«ical actiocLB] RIQtn-ADJilCEVT: 

(AUmiiAT-DiTTirE) 
d4tA^ SYLLABLES: 1 

catttfiojy, liUXILlAnV-DBJECT 

7'- canegtjry: VEElfl 

^iLdt«ti: 01 da^A- CON JUG ATI DN: 3 
nn rpliQBfl . PTTA 
CMiflSorj: VEJtB 

l: itta- COHJUGATIDff: 3 
j&Drph&nfi: ETYI 
&atHg[>rr: TEfl^E 

SS: projection?! Nil, 

iiateg^rj: ALTULIAEir-BI^B 
chtldMIi' proidcticin?: HIL 

data: AEHJUnfiHT^ verb 

category: VEHB 



etlliriiiE projection?: NIL 

data: TETETA- 1,53 1 Gtf E& ■ laEMT 

ClL5E-Al55rGBEDr EllGliTlVE 
cat&^ory. CASE 
cliili3r*Ii' proJectioD7r HIL 
Hi4»^ PERSflU: 1 

IfUnSEFL- (SSAGULIlEIJ 
nchTph«&a[ KGAJULU 
cat^iorj: BOUB 

ptCHj*ction^ : T 

d&ta: CASE-Hj^EiKED L EFlC^ATIVE 
KavptiQne : ALU 
catflgpry: CASE 

projection? T T 
category ^ VERB 
chtldrfiD. pniJ*Ctlt?n»: T 

iCtlons: CiSE-JiSSIGB; ABSDLUTIVE 
THETA-ASSTCtf: THEKE 

Kiata^ -KflSE: MOITPIST 

THCTA-MLE3: (AfiENT THEME) 
kQtHTi EEC ATI VE 
THEKE. AB^niimUE 
SimJECT- AGENT 
OBJECT" Ti^EHE 

b^ri^fPia^ NYA 

cBtft^oty; VERB 

|EroJBctic?Q*r T 
Clt4gery- AUIILI1EY-B4SE 

flhildrfln: proJoCtlctii*: NIL 
data- PEP^QK. 2 

nimilER SItfflULAR 
nurphene VQEU 
catflsoxj: AUS] Li ABY- OBJECT 

pr^njoctton': HIL 
data- PEJ^aEf: 1 

miMBER- SIEf4lULAR 
Qu^rphaniq : BtfA 
citafiory; AUlIUARY^StfBJECT 

proJflctlBn?: T 

data: TEKES: (UnHPlST) 



ise 



IISPECT: IWEBFECt 
catBgOT,; lUJILTAHV-flflaE 

A. 2,12 Tdd Mnny Argument" 

CCafl VI KA>] UXUPDU m)) ((Rmi))) S- «P6r.*maxic*J. 
Thfl syptactlc BCEUCtora U upcumi-c tad . 

PS: 01 catBBory: VERB 

chiUrao. 0: d*ta: CONJUGATCQH: £ 

li iatBT CTDUUGlTIQi: & 
■orpFiame: HI 

1. l«Ua3 -,Cti.u^. mTLUItY-EELECT: AlTULl ARY-DATIire 

ACimiAFL^-SECECT, lUIlLURY-OBJECT 
lUIILrABY- SELECT; AlHILI m- SUB lECT 
KlflHT- ABIi-CEBT : ( iUHLl IHY-SUB JEtTT 

iLUXILIiJfY-DBJ^CT 

IDHLUnl- DATIVE) 

data: SYLLABLES: I 

rurphDnB: KA 

caMfiory; iUJLl LIIHY-BISE 

2' phrasal BctiP^a: SELECT-: KDOB 
cat.Boij: CASE 
^hlldxan: Or Diarphwma: KITMJU 

citagPFV' ''°'^'' 

I: Bi^rpbano: 'AHS" 
catsaory: CIS-E 



157 



SS: projection?: IJIL 

categcsry: AUI ILTAEIT-H JlS E 
cbiildr»n: prtjactiDnT. NIL 

data: AHCLmEKT: VERB 

cA?4^ryj YE3.B 
chLldr^nJ prdJactLon? . iflL 
I data; THErii-ASSrCSED: THEfl£ 

CA3E-JS3HjRED: ilB5&LlJTIVE 
catflgi^rT" CJiSE 
ihildranL pro j act ton?: IfIL 
Hh^TpLaiia : KUJIHU 
categcn^. KaUH 

projection?: T 

dAta- CiSE-EUIHXEQ: ABSDLUTIVE: 

morphenar 4^$^ 

C^tagar;: CASE 

prf>jactinn^ : T 
citejory; '■'■EftB 

Childi^en; pTO]acticin7. T 

d4ta[ TEEfSE. N^NPA^t 

TUETil-TtDLES : tTBEHEj 
TBEHE^ RBSOLUTIVE 
SUBJECT r TffEHE 
piQrp^.ana: YA 
catagpry: VEHfl 

pro j ecTlQn? : T 

c atego rj *UI1 LI XR Y-B ASE 

ttilldran: pro j a c 1 1 i?ft^ T 

d^U- TEJISES: (TJOHPA^t} 
ASPECT^ IKPE3irE{:T 

ta&rpftftfflfi^ KA 

catagorj: AUILLlABi-EASE 

A*Z*13 CEUiO Marking 

Pwsizig (tUlL lyafl H^rtU f^AJl ((YUUlA Kl>> ((HGAJULU}] 
((HYlTNTirLU KU))). 

FE: d- catefiorj; ftUU LI ARt -OBJECT 

chlUr^n: 0: data; SY1.U^LES= 1 
norphaoie' HA 
cai:e^rj. iUTLlLIATtV-QABE 



IH 



1- data; SVLLIBLES: 1 

CBt*aory: AUl ILI ART-SITB JECT 

2: dttiL STLLABLES: i 

catafiorji AUITLliFL'f-DBJEtTT 

3t data: SYLUELESl 1 

catflgc^rj" AintlL^lRY-DATlVE 

chilHlr*fli 0: dB^4 CCBJUGATIDJI^ £ 
^KtMg^rg-- VEflB 

1: dttEL CO]IJUCATIDfl^ 1 
niorph*ffl* - HI 
catoK^ty; TEM3E 

2- phrasLL aCtioBBi SELECT*! BOtI* 

childmi' 0[ nrorphftina: BCi-TUtU 
CAtfigory: JfOUB 

1 1 BifirpheiBB - *AB5* 
catfl^Gry: CASE 

3: phrasal actlpin^ SELEO^ BDUa 
CBtepiry- CASE 
childran- 0" iiorphBBin liYUimJLlJ 

1 ; BOrpllfiDB t KO 

{^atagory: CASE 

5E: prQJflctii3TiT' iflt- 

catoB^ry^ lUIlLTAHY-BASE 
chJ-ldxen. projactiph'" JfIL 

data. ARflUHEHT. V£Jtfl 

category; VERB 

childr4h^ pnjflCUon?! NIL 

dsTia: TH&r*"ASSTaJlEn" THEHE 

CASE^aSSlCNED: IBSDUmvE 



IJU 



children [ ^'^^J^^'^^^^^^ ^^^ 
d.Kta; PERSDPf. 1 

EfifflBER- (SrHGlTLilU 
nKiTphQoie : Ef^A JDLU 

data; CJlSE-HAFLICZD^ IHSDLUTITC 
B^rpbeme: *AB£+ 
category^ C^E. 

prgjectiony T 
categDjy: V~QtB 

drta: THETIi-UEiKED. PATH 

NUMBER: CSrMCULAR} 

Kirph^b^ rmjEnuLU 
category . ff^UIT 

pTO]acticm^i T 
4»a: CASE-.lSE1GnEDi 
TATrVE 
CASE-MARtLEllJ ■ OxTlVS 
THETA-ASSIGXED: 
PATH 

cat&gory- CASE 

prD]ectioDT: T 
data; T^2E^ taOXPI.fT 

TKETA-aaLES- (TEIEHE PiTJJ} 

TTIEME: ABSDLUTITE 

Pfi713; DATIVE 

SUBJECT r THEHE 

OBJECT: PITB 

G-ate£ory: VEAB 



childraD: pro J action? j NIL 



in 



data. PERSnF. 2 

nnrptiEiu^ NGKU 

categurj" JlIH^I LI AltY-OH JECT 

pTOJBCticitl^ ■ UIL 

dtta- PEIL^^Pn- t 

tatflftory. fcUlILIftElV -SUBJECT 

pro]ecti*ipT: T 
da^*- TE]fSE3- (HDHPA3T) 
|,5P£CT: IflFERFECT 

catflft^ry^ ^UJLTLIXFLT-BliSfi 

ParBlng *(((£! RKA HGKU MJiJ) ({YULEA Ml)) (1HC*JULD ALU)) 

[(fXA AHA HGKU JILJl)) C CYULItA EII» <(SGUULU HLU)) ( (FfUXTULU » ) .1 
Tfia Bjntacttc Etructor* Sfl EmccHrLTta^t ad . 

C((J:a ma tfCXU HU)> HYULU Kill (U^^*™-^ KU" KHYUlfTULU FLLU)» if 

Thi syntactic ■tructura iri uriconDact^d- 

A. 2-1-1 Auxiliary Baec Agrefiin^ent 

PaxBiofi .(ClNGAmU iltU LPl m BUH ((P^TUTA RNTll HKJMU KUJl 
UEIiiail)). 

((lUGAJULU RLU LFA iUtt FlLAM aPUSTA HHI)) CCKUKDU XU)1 l(XAnLlJ» i- 

ucgramB't ^ caI . 

lb* tanaoH of LP* and PUHT* do not njtch. 

PitBlra CCillCimU RLU LP* FIH* RLil)! iiPIffVt WJlDl (CKURDU XU)} 

lEil 



FS: 0. pbiras^ acticnu: SELECT*: HOUK 

children- 0- marphdinfr^ HCJlJULU 
catQgQrT' HDU}] 

taHflgOFT. CASE 

1: CftlQ^or?^ iUIILIilHT-SUBJECr 

thildniii Ct Ifl^ltit Jittionffi AtrarUART'SELECT: 

AUULIUlV-QBJECT 
dBta^ SYLLiSLES^ 1 
■□rpbflmai LPi 
category- iUirLrAFLy-BASE 

1] dita- SYLLABLES: 1 

catafiory- AUII LI AFlT-SUB JECT 

2; dBtnr SYLLABLES: 1 

categpryr AUirLIARY-SATrVE 

chlldxHn: 0: data; CQVJUCATIOH. 2 

1^ tiAta: CDirjr;cAriDK; 3 
narpheiBe: RHU 
tatagiiryi TEWSE 

S: phraaal ^LCtions^ SELECT^- HQrJV 
CAt4£cry: CA3£ 

cbj.IdjBD: 0: oiDrpbeiDfr: }[UFLn-U 

cBtagdryj CASE 

4: phraaal actions: SELECT^: NDUtJ 
catfigtjry: CA5E 
ChildtQQ; Q-, Borptifian: lURLI 
catflgory: TJOUH 



MB 



1- marphBai. *ABS* 
Cfti^dgo jy : CiSE 

SB: prfljaction?! VIL 

Childraa: prflJflC^lon"^ ' NIL 

data[ AUGlTOafT: VEfLfl 

cat4e*iry^ VEItB 

chtldj^n: projsctionT: ECIL 

data^ THfir*-ASl5IGKED: A^EUT 

CASE-ASSHSNED: EJtOATIVE 
catBgor/. CASE 
dilLdranr pro J oc ^ l-ii:' ^ ifIL 
dhta] FEASDN. 1 

KVBGEl: CS£T3CjULAR) 
Eutrphwnv: jai^lijum 
C*t>£eryj HOUV 

dfttfi- CiSK-HUU[EI>: EfU^ATlVE 
DiDrpbama; RLU 
CAT-A^ary: CASE 

projattion?: T 
category: VEKB 

ctLLIdran: prEi] flctl^Ki? ^ JUL 

flataL THETA-liSSICEfED" TKEME 

CilSE-iSSICNED: ilBSDLUTTVE 
catagoxy. CliSt 
ohildran; projection?: NIL 
BurpheJDfii^ KARL I 
CtttgChry: MOUW 

pr&JattiQn?: T 
data: CASE -HARKED [ 
ABEDLUTIVE 
Hcrphfloe. •ABS* 
catflg^ry^ CASE 

ptcj action?: lill 

ditt. TKRrA-LIHKEO: PATH 

cbildren: pre J action^ : fflL 
[D^rt^asLe^ KURDU 
category: HOUW 



IftJ 



proJficti&Q*' T 
4ttt.: CASE- 4^&IC SEP; 

CASE-HAHKEI); 

DATIVE 
THETL-ASSICIiED] 

PATH 

dataj TEHSE] FAST 

THETA-nOLES; (ACENT THEHE 

PiTHj 
MEHT: EflClTTVE 
THEHE: ABMLUTTVK 
PITB] DATIVE 
SUBJECT- Al3E3fT 
HBJECTr PATH 
unrphBuaj PUHTA 



ca^EfiQi-y: AUSTLIlItT-BlSE 
ctiildtfln^ projfictl-oii*^ HJL 

mo jptiQUB ' FlLA 

Citeg^ry. iUXILrAflV-DATIVE 

pro j action Tt CCIL 

laOKBZn^ 2I](CULAEl 
■OFphaDLft^ iOlA 
CAtftj^ry: iUi:UABT-3tJBJECT 

pr?J*cti an"^ ^ T 

4at&^ TEEfS?^^ (PfST IFlRtALISl 

ASPECT! rUPEftFECT 
BDiTihBDLe I LFA 

c&L*Ei.ryi juiiuarLY-BAgs 



f«iiiag "(((KA RBA ItGHtl HLA)) ((TULKA I1I» ((HAE11U)J 
((JTflFNTULU KIT))). 



U34 



usgrumiati i;^! - 

Th* AUJlLIAnY-StfEJKCT clitic floifl not agr?* in parBon tflth ^he 

SUBJEITT- 

ParBlhg •<((K1 UKA R5Kt| TtLAjJ ((YULM Mr>) C(HCAJULU)) 

(((KA RMA KKU HLiJ) [(¥UU(A HX» ((HGUOLU)) ((YIREilflJI XX>» is 

The lUXILIAKr-OflJECT clitic d&fia nQt agr^* In person ifith thfl OBJECT 



1B£ 



Bibliography 



jAljftST] S A busy- LkenEiaR and Par^r^. In JV£^i5 /ff^ Noith l^astem LLn^iFtk 
Society^ Univen- i l>' i>[ MaflWtcJntfletts, Amiierst. MA, IflSV. 

iBrirS? J. Bti]=sn»rL. Tie Afen^n/ Wcj^reflcnJa^cQn ^^f Grammt^tital nWdhnni?- MIT 
Prefis, Cimbnd&c, MA, 19B2. 

|BWfi4] R.C B^rwicL AQcL A.S- Weinherj Tftc GroFrLr/iafiira/ fffliis y/ t^jyu*.-*- 
tit p£rjcrfY}aru^: Zdn^dgc E/sc OFirf Ar^i^^Pflrt. Volume || of Current 
SluJcM in Lin^^stic^ MIT l^r&R^ Cambridge, MA^ L!>e4. 

jCln?65] Nr Choni&ky Af^p^U of ih^ Thcorj^ of Ssfntoj-. MiT Piffl-^, CwTlhrids^^ 
MA, IM.^- 

[Clio3l| N. Chomsky. tftr(i**fl* on G^JtiemFnenf afid fiandtn^r ^^^ '^^^^^ I.«t(tJJT^h 
Volume 9 of ^Itidces in Grncmfiiit Grornmari l^'oriB PubJlCfl^JtlIl^^ Dor- 
drecht, isai. 

|Cl]0S2] N. Clionwky Sr^n^^ Ctmcepts and ConHgrjcnretf d/fJjr TfieorJ/ o/Gflfem- 
mcnl DFcJ EfHrffrtyr VdunifcC of tujyi^'J^^'^ Inquiry M^nuffvuph^ MIT PrrsBf 
Cambridge, MA, I9H3. 

IChoSEi] Nr CliO]S3Sty. Kn^turfr^ge r>/ Ij fi^uajje ■ /Iff fl'alurc, Dri^n, ond [/Mr CrJrt- 
irciTjcTwrc, Prac?c^^ WftStpOrt, CTi 1980, 

|Cur6[] H-B_ Curry- Socue Lopcai x^pwLs of Gr^minfttkrAl Ht-niftmfl- In Hn 
Jakobaon, j^dLlof, iVr^rCvrt j^/ tfltufyapf and lis MathcuiGiicQl A^pcct^^ 
Proceedifiqf^ i?/ the Twelfth Sj^mpfrvn^m m Ap^tf^d jTVcif/ifffta^tr_*, pagea iLC- 
63, American MitllOtnatLCft! Swwty, proindencct Hli IBBL. 

[Oora7j a. Dorr UNITItAN: A Prrtx^-ipte- Bitted Approach to hTuchin^ TVanrfa- 
iion. Ma^ter'n the&i^i Masflachu&eLta InalitnlE of T^^lmLi Injgy, I9S7- t'dftb- 

[EdidTG) J. EEiHrade. A JVnrtff/tfnnilforwJ /pjifMcfi tn h'ry^ii^h S^fninx, Acadenuc 
Preaa, Lcuidon, 137& 

[Hal93| K. H41IU. \VaTl[HLri ajid likC Gnu^sniar of NoTi-conflgii rational JnanguaHSfl 
PrrHunii /.a^ljpi^a^ ind fAn^i^tic Thecrff^ L5-47t lOFU. 

1«6 



EJo!lS&| M JohnBon. Pairiiig with Dmorthnijiifi CoufltitucJitS. Id 85>^ j^nnnu/ 

32i Ae^cminatjoii lor CtiinpijUttl<ill&l l.lriSutEtkfi, iy95. 
IJc*] A. JoBhi. The ConverftPT^re of Mildly CiMHtXt-WUsHlvE ForraaljBniE (or 

|K]Afl2] J. XliViiis, ConliguriLLon Ln Non-COtifL^uratlonal LanjuaRCT. In Fr<J- 
efa=dtng? of th^ HVri Coasl CcJK|/fran££ on Fonjiat Lm^l^Uc^^ Stanford 
UnaverHity, t>t(Liirord, CA^ 1^82. 

[LanTal M- Laiighren DirrttJOJLlI Terminology m WaHpiri, A Otntral AiLflLialLii] 
Lin^i^C, IVtrtin^ Papers in tartpudffe -Jfl-' tin^ulicflj 9rL-l6j I97B. 
Ta^manh^n College of Ad^antcd EclutfttlOFl, l-aunceskin. 

[LcvflS] B- LevLFi, editor. L^TtOlt S^tmul^C^ in fleiiiciu, CtntnM for Cognitive Scl- 
encti, Uc*4¥4JjJli|i^l9 lnatittile of THKhnolf^gVH ]9P^- 

|MarS2] D- Maix. i^Hion. W.H. Fryman md Corapaoy, SaJi FrancUco, 1^92- 

[Na£86j D. Naflh. TcJ/hh-J *n Warrpin" Grtu^ntar- OutfiianJcng Du^crCdli^^iU in 
Zi'n^fjJiVs^ Garlind FuhlnliJPg lnc_. New Yorkn ISSGr 

{SbiiSS] JLM- Sharp. 4 Model oJGrGmmarB^^^d^f principles GfGouer^memoTid 
BiTulins, HasU'r'ft thefljft. The I'luveraily of Btiltsh Colmnbla, October 
19£5. 

[S]iiiS3| J.H. Slmpaoil- H^i*/!^:!.^ of ETaFVpin Mcprp^oJtptfj (i"d ^jniai- PliD Ihcsii, 
MafisaclLQBHiLLs Inilltutc of T eLhnoloj^y, Apnl 1^83, 

[StcBl] T.A. SloWiJl. OrjiJifi.9 of f^hruM .5inic(ijrc. PIlD tlieHlft, MaaBiclm^etta 
Inatjtiite of Technology, Carnhrldge.M At September 1981, 

|vR^V8G] H, v^n «JemfidLjk aod E William^, hnrttdurtim to th^ TftfOPj of Cram^ 
fnar. Volume 12 uE C-'liJTf fH n^f u Jaefl I'n tinffuwKV^n MVt PreaB^Ciinhridfir?, 

[Wacyj[ J. Wiiktmas^l. Ob^r eiii tfcCBeti dci IJidogprmflniKlieEi Wor^sHiUnftg- h- 
Atofff^nuouiBche For^^h^^n^en, 1:333 -I^^t 1892. 

lWeh84l E. WehrlLr -1 Gov^mm^ni^BiwuHng Poffltr >r fVcnc^. Tftthrtafil Re- 
port -13, UnlvenjCe de Gen^v^, 1934, 

[Wilei] E. ftnlltfLm&- Arpimem Structnirt and MorpboloE^. Th^ f.iTijtiisltc fte- 
fiew^ 1:81-1H, I98L 

[WllBJ] K- WlUiami. Graanmatical R^Uticiia- Li^^ui^Hc Ift^tr^, lb{4y£1^a-U, 
Fall I9B4- 



IBT