Resolving Zero Anaphora in Japanese

Resolving Zero Anaphora in Japanese. Tadashi N o m o t o and Yoshihiko N i t t a Advanced Research Laboratory, Hitachi Ltd.. 2520, Hatoyama, Saitama 350-03, JAPAN E-mail: {nomoto, nitta}@harl.hitachi.co.jp. Tei.+81-492-96-6111 Fax. +81-492-96-6006. A b s t r a c t . The paper presents a computational theory for resolving Japanese zero anaphora, based on the notion of discourse segment. We see that the discourse segment reduces the do- main of antecedents for zero anaphora and thus leads to their efficient resolution. Also we make crucial use of functional no- tions such as empathy hierarchy and mini- mal semantics thesis to resolve reference for zero anaphora [Kuno, 1987]. Our al)proach differs from the Centering analysis [Walker et al., 1990] in that the resolution works by matching one empathy hierarchy against another, which makes it possible to deal with discourses with no explicit topic and those with cataphora [Halliday and Hassan, 1990]. The theory is formalized through the definite clause grammar(DCG) formalism [Pereira and Warren, 1980],[Gazdar and Mellish, 1989; Longacre, 1979]. Finally, we show that graphology i.e., quo- tation mark, spacing, has an important ef- fect on the interpretation of zero anaphora in Japanese discourse.. 1 I n t r o d u c t i o n . Over the past years, schemes like Focusing and Cen- tering have dominated computational approaches to resolving anaphora [Sidner, 1983; Walker et al., 1990]. Their success derives from the utility they have in identifying salient discourse entities such as topic and thereby locating the antecedent for an anaphor. But they all suffer from the problem of. directionality; they process the text (the list of sen- tences) from left to right, picking out focus along the way and see if a anaphor corefers with a focus already encountered. W i t h the one-way processing, forward-looking pronouns (cataphora) are not pos- sible to resolve. Since Japanese has great tolerance with forward reference, a proper theory of zero pro- nouns should meet the problem of directionality.. In what follows, we discuss some points about dis- course segment and zero pronoun in Japanese. We begin by introducing the idea of discourse segment. Consider the pair:. (1) Taro-go sara<i>-wo dasi, Hanako nora plate ace prepare-and. -go 02<i> ryori -wo morituketa. nora food acc arranged. Taro prepared the plates, Hanako arranged food on them.. (2) Taro -ga sara<~> -wo dasi, Hanako<i> -wa top. 01<i> 02<k> ryori-wo morituketa.. Taro prepared the plates, Hanako arranged food.. Here, 02 represents a suppressed expression. It acts as an indirect object of the verb moritsuketa. 1 1 and 2 are morphologically identical except that 1 has ga (nominative marker) where 2 has wa (topic marker). But they differ widely in meaning:l im- plies that Hanako arranged food on the plates that Taro prepared, the reading 2 does not imply; in 2,. 1Here and throughout, we intend the term 01 to rep- resent a zero pronoun for the subject, 02 for the indirect object, and 03 for the direct object.. 315. H a n a k o could h a v e a r r a n g e d f o o d on p l a t e s s o m e - b o d y o t h e r t h a n T a r o p r e p a r e d . Now l o c a t i n g t h e difference will involve t h e n o t i o n o f discourse seg- ment. A discourse s e g m e n t is defined as a set o f sentences which a p p e a r in s o m e region o f t e x t a n d which is d e l i m i t e d b y a t o p i c - p a r t i c l e wa. T h u s 2 b r e a k s u p i n t o two s e g m e n t s , a clause w i t h Taro-ga a n d one w i t h Hanako-wa;1, c o n t a i n i n g no w a - m a r k e d e l e m e n t , f o r m s a s e g m e n t b y itself. Section 2.1 pro- vides s y n t a c t i c definitions for t h e discourse s e g m e n t . . A n o t h e r i m p o r t a n t f e a t u r e o f discourse s e g m e n t is t h a t o f c o m p l y i n g w i t h t h e M i n i m a l Semanti cs The- sis ( M S T ) [ N o m o t o , 1992], a f u n c t i o n a l p r o p e r t y t h a t m a k e s a s e g m e n t cohere. T h e M S T says, ' A s s u m e as identical any pair o f zero pronouns i f it is p a rt o f some segment and does n o t occur as argu m en t s f o r the segment's predicate.' T h u s a n y p a i r o f zero p r o n o u n s t h a t fall into t h e d o m a i n o f discourse seg- m e n t are t a k e n t o b e coreferential, unless t h e y occur for t h e s a m e p r e d i c a t e . 2 Significantly, t h e M S T is a m e n a b l e t o s y n t a c t i c t r e a t m e n t . . I n a d d i t i o n , we m a k e use o f ~he empathy hierarchy t o choose b e t w e e n coreference r e l a t i o n s h i p s a d m i t t e d b y t h e M S T . W e specify a p r e d i c a t e for t h e e m p a t h y h i e r a r c h y a n d resolve zero a n a p h o r a b y u n i f y i n g one p r e d i c a t e ' s e m p a t h y h i e r a r c h y w i t h a n o t h e r which occurs in t h e s a m e s e g m e n t . Since unification is a n o n - d i r e c t i o n a l o p e r a t i o n , we are able t o t r e a t for- w a r d as well as b a c k w a r d reference.. 2 T h e o r y 2 . 1 G e n e r a l . A discourse s e g m e n t (DS) is a t w o - p a r t s t r u c t u r e consisting o f h e a d a n d b o d y ; a h e a d is a n o m i n a l w i t h a wa m a r k i n g ; a b o d y is a set o f sentences, which e n d w i t h a p e r i o d . N o t e t h a t a n a d j u n c t i v e clause is n o t a sentence here, since it ends w i t h connectives like .node because, .kara because/after, .to and-then, etc. F o r m a l l y , we a s s u m e sentence h a s t h e following analyses, which a r e given in t h e D C G f o r m a l i s m . . (3) S -> C+, N(pp:1~a).. S -> C*, N ( p p : ~ a ) ,C+.. S -> C+.. c+ d e n o t e s one or m o r e occurrences o f clause, C* zero or m o r e occurrences o f clause, a n d N ( p p : wa) denotes a w a - m a r k e d n o m i n a l ; p p : w a specifies t h a t t h e a t - t r i b u t e p p (for p o s t p o s i t i o n ) h a s wa for t h e value.3Let us define discourse s e g m e n t by:. 2 [Hobbs, to appear] talks about the cognitive economy in understanding discourse: it says in effect that coher- ence is the result of minimizing the number of entities in discourse.. 3We take a wa-marked nominal to be a sentence adver- bial. Thus our approach differs from the tiaditional gap analysis of topic construction [Kuroda, 1965; Inoue, 1978; Kitagawa, 1982; Gunji, 1987], which assumes that a wa-. (4) D - > S+.. a n d t e x t b y . (5) T - > D+.. As discussed in section 1, we choose t o r e s t r i c t D t o c o n t a i n i n g a t m o s t one ~1 ( p p : w a ) . W e i m p l e m e n t t h e restriction b y way o f s o m e a d d i t i o n s t o t h e rule set 3.. (6) a S(head:X) -> C+, N(morph:X,pp:wa). b S(head:X) -> C*, N(morph:X,pp:,a),. C+.. Here, t h e 6 rule t a k e s care o f i n v e r t e d sentence a n d t h e 6 rule n o n - i n v e r t e d sentence. T h e rule set 6 enforces unification b e t w e e n t h e h e a d value a n d t h e m o r p h value, m o r p h r e p r e s e n t s t h e m o r p h o l o g y o f t h e n o m i n a l ; t h u s m o r p h : t a r o specifies t h a t the associ- a t e d n o m i n a l h a s t h e m o r p h o l o g y " t a r o " . . Notice t h a t unification fails on a m u l t i p l y h e a d e d s e g m e n t . A h e a d a t t r i b u t e , once i n s t a n t i a t e d t o s o m e value, will n e v e r unify w i t h a n o t h e r . Unifi- c a t i o n , therefore, a c t s t o l i m i t each s e g m e n t in t h e discourse t o a single h e a d . N o t e also t h a t a n n o n - h e a d e d discourse, t h a t is, discourse w i t h n o h e a d e d s e g m e n t s , h a s a l e g i t i m a t e DS analysis, for unifica- t i o n is possible b e t w e e n e m p t y heads. T h e following lists t h e rules for DS G r a m m a r . . (7) T -> D+(head:_).. D(head:X) -> S+(head:X).. S(head:X) -> C+,N(morph:X,pp:wa).. S(head:X) -> C*,N(morph:X,pp:wa),C+.. S(head:_) -> C+.. 2 . 2 H e a d e d v s . N o n - H e a d e d D i s c o u r s e . T h e discourse c a n h e p e r f e c t l y intelligible w i t h o u t a n explicit t o p i c or w a - n o m i n a l , which i m p l i e s t h a t a discourse s e g m e n t m a y n o t b e h e a d e d a t all. I t a p p e a r s , however, t h a t a discourse s e g m e n t a l w a y s c o m e s o u t h e a d e d e x c e p t w h e n t h e r e is n o h e a d avail- able in t h e t e x t . I n f a c t , a s e g m e n t a s s o c i a t e s w i t h a h e a d n o m i n a l regardless o f where it occurs in t h a t s e g m e n t . . (8) T a r o < i > - w a 01<i> 0 2 < j > seki - w e u z u t t e top s e a t acc give. - a g e t a node, 01<i> 0 2 < j > orei - w e help b e c a u s e t h a n k . i w a r e t a . Ol<i> c h o t t o t e r e k u s a k a t t a . say pass sl i g h t l y e m b a r r a s e d cop. nominal is dislocated from the sentence and leaves a gap behind. In fact the analysis meets some difficulty in ac- counting for the wa-nominal having semantic control over a set of period-marked sentences. cf. [Mikami, 1960]. Ours, however, is free from the problem, as we see below.. 316. Because Taro gave h i m / h e r a favor of giving a seat, he/she thanked Taro, who was slightly em. barrassed.. (9) 01<i> 02<j> s e k i - w o u z u t t e - a g e t a - n o d e , T a r o < i > -wa 01<i> 02<j> orei-wo iwareta. 01<i> c h o t t o terekusak - a t t a . . Because Taro gave h i m / h e r a favor of giving a seat, he/she thanked Taro, who was slightly embarrassed.. (10) 01<i> 02<j> s e k i - w o u z u t - t e - a g e t a - n o d e , 01<i> 02<j> orei-wo iwareta. Taro<i > -wa 01<i> c h o t t o terekusak - a t t n . . Because Taro gave h i m / h e r a favor o/ giving a seat, he/she thanked Taro, who was slighau embarrassed.. 8, 9 a n d 10 each c o n s t i t u t e a discourse segment headed w i t h Taro. 4 A discourse can be acceptable w i t h o u t a n y head at all:. (11) 01<i> 02<j> seki wo u z u t t e a g e t a node, seat ace give favor because. 01</> 02<j> orei -wo iwar eta. 01<i> t h a n k s ace s a y pass. c h o t t o terekusa k a t t a s l i g h t l y e m b a r a s s e d cop. Because he/she gave h i m / h e r a favor of giving a seat, he/she thanked him/her, who was slightly embarrassed.. T h e speaker o f 11, or watashi I would be t h e m o s t likely a n t e c e d e n t for t h e elided subjects here; who- ever gave the favor was t h a n k e d for the kindness. Let us say t h a t a discourse is headed if each o f its segments is headed, a n d non-headed, otherwise. O u r a s s u m p t i o n is t h a t a discourse is either head ed o r non-headed, a n d n o t b o t h (e.g. figure 1, figure 2). 5. Formally, this will be expressed t h r o u g h the value for the h e a d a t t r i b u t e . . (12) T -> D(head:empty).. An empty-headed discourse expands into one seg- ment; its head value will be inherited by each of the S-trees down below. Note that unification fails on. 4The Centeringalgorithm is not equipped to deal with cases like 9 and 10, where the backward-looking center Taro refers back to an item in the previous discourse.. sit is interesting to note that a multiple-head dis- course may reduce to a single-head discourse. This hap- pens when discourse segments (DS) for a discourse, share an identical head, say, Taro and head-unifies with each other. In fact, such a reduction is linguistically possible and attested everywhere. Our guess is that a repeated use of the same wa-phrase may help the reader to keep track of a coreferent for zero anaphora.. T. /\ D D. I I sl. Figure 1: Unacceptable DS-tree. "S O" denotes a sen- tence with a wa-marked nominal.. T. I D. /\ sl. F i g u re 2: Acceptable DS-tree. the head value if any of the S's should be headed and thus specified for the head attribute.. The following rule takes care of headed construc- tions.. (13) T - > D + ( h e a d : . ) . . The rule says that each of the segments has a non- null specification for the head attribute.. 2 . 3 M i n i m a l S e m a n t i c s T h e s i s . M i n i m a l S e m a n t i c s Thesis ( M S T ) concerns t h e way zero p r o n o u n s are i n t e r p r e t e d in t h e discourse seg- m e n t ; it involves an em p i ri cal claim t h a t t h e seg- m e n t ' s zeros are coreferential unless considerations o n t h e e m p a t h y h i erarch y (section 2.4) d i c t a t e t o t h e contrary.. (14) Kono ryori<i> wa saishoni 01<i> mizu this food acc first water. wo irete kudasai. Tugini 01<i> sio acc p o u r in i m p e r a t i v e n e x t salt. wo h u r i m a s u . 5 h u n sitekara, 01<i> ace p u t - i n m i n . a f t e r p a s s i n g . niku wo iremasu. m e a t ace add. A s for this food, first pour in some water. Then put in salt. Add meat after 5 rain.. We see t h a t 14 c o n s t i t u t e s a single discourse segment. According t o t h e m i n i m a l sem an t i cs thesis, all o f the zeros in t h e seg m en t are i n t e r p r e t e d as coreferential, which is consistent w i t h t h e r e a d i n g we have for the ex am p l e. Here is a m o r e c o m p l e x discourse.. 317. (15) T a r o - w a 01<i> m a c h i - n i i t t e , 01<i> huku top town to g o cloth. - w o k a t t a . Masako<j> - w a 0 1 < k > sono acc b o u g h t top t h a t . huku -wo t a n j y o b i -ni m o r a t t e , 01<k> cloth acc b i r t h d a y on g o t . t o t e m o yoroko - n ' d a . m u c h rejoice p a s t . Taro went downtown to buy a clothing. Masako got it for her birthday present and she was very happy.. T h e first two zeros refer to Taro and the last two refer to Masako. B u t this is e x a c t l y w h a t the M S T pre- dicts; 15 breaks up into two discourse segments, one t h a t s t a r t s with Taro-wa and the o t h e r t h a t s t a r t s with Masako-wa, so zeros for each segment b e c o m e coreferential.. 2 . 4 E m p a t h y H i e r a r c h y . It a p p e a r s to be a fact a b o u t J a p a n e s e t h a t the speaker o f an u t t e r a n c e e m p a t h i z e s or identifies m o r e with the subject t h a n with the indirect object; a n d m o r e with t h e indirect object t h a n with the direct o b j e c t [Kuno, 1987; K u n o and K a b u r a k i , 1977]. In fact, there are predicates in J a p a n e s e which are lexi- t a l l y specified to take an e m p a t h y - l o a d e d a r g u m e n t ; yaru give a n d kureru receive are two such. For yaru, the speaker e m p a t h i z e s with the subject, h u t w i t h the indirect object, in the case o f kureru.. T h e relevance o f the speaker's e m p a t h y to th e reso- lution p r o b l e m is t h a t an e m p a t h i z e d e n t i t y becomes m o r e salient t h a n o t h e r elements in the discourse a n d thus m o r e likely to act as the anteceden t for an a n a p h o r . . (16) T a r o - g a Masako<j> -ni hon -wo k a t t e n o m to b o o k acc b u y . - k u r e t a . I m a d e m o 01<i> sono hon -wo h e l p e d still t h a t b o o k acc. daijini siteiru. care k e e p . Taro gave Masako a favor in buying her a book. She still keeps it with care.. In 16, 01, subject o f the second sentence, c o r d e r s with t h e indirect o b j e c t Masako in the first sen- tence, which is assigned e m p a t h y by v i r t u e o f t h e verb kureta.. Formally, we define the e m p a t h y hierarchy as a f u n c t i o n with three a r g u m e n t s . 6. empathy(Z1, Z2, Z3). 6The definition is based on the observation that Japanese predicates take no more than three argument roles.. W i t h t h e definition at h a n d , we are able t o f o r m u l a t e t h e lexical specification for kureru:. V(empathy(hrg2, Argl, Arg3), subject : hrgl, obj ect2 : Arg2, object :Arg3) -> [kureru].. yaru has t h e f o r m u l a t i o n like t h e following:. V(empathy(hrgl, Arg2, hrg3), subj o c t : hrgl, obj ect2: Arg2, object :Arg3) -> [yarun].. F u r t h e r , let us assu m e t h a t variables in t h e em- p a t h y h i erarch y represent zero p r o n o u n s . I f a vari- able in t h e hierarchy is i n s t a n t i a t e d t o so m e non-zero i t em , we will rem o v e t h e variable f r o m t h e h i erarchy an d m o v e t h e i t e m s following b y one _position t o the left; we m i g h t call it empathy s h i f t i n g / Now consider t h e discourse:. (17) 01</> 02<i> h o n -wo y a t t a -node, b o o k acc favored because. 01<k> 02<a> orei -wo iwareta. g r a t i t u d e ace s a y cop. 'Because he/she gave a book to him/her, he/she was thanked for it.'. (18) a e m p a t h y ( 0 1 < i > , 0 2 < j > , _). b e m p a t h y ( 0 1 < k > , 02<9 > , _). 18(1) corresponds t o t h e e m p a t h y h i erarch y for the first clause in 17; 18(b) co rresp o n d s t o t h e hierarchy for t h e second clause. Unifying t h e two s t r u c t u r e s gives us t h e correct result: n am el y , 01<i> - 01<k>, an d 02<i> = 02<9 >. Notice t h a t zero i t e m s in t h e segment are all unified t h r o u g h t h e e m p a t h y hierar- chy, which in effect realizes t h e Mi n i m al S em antics Thesis. As it t u r n s o u t , t h e M S T reduces t h e n u m b e r o f s e m a n t i c a l l y distinct zero p r o n o u n s for a discourse segment t o a t m o s t t h r e e (figure 3). We conclude t h e section w i t h a listing o f t h e relevant D C G rules.. / S ( e m ~ Z 3 ) ) S ( e m ~ Z 3 ) ). F i g u re 3:. D(head:X) -> S+(head:X.empathy(Z1.Z2,Z3)). S(head:X.empathy(Zl,Z2.Z3)) ->. C+(empathy(Z1. Z2.Z3)), N(morph:X.pp:,a).. S(head:X,empathy(ZI,Z2.Z3)) -> C*(empathy(ZI,Z2,Z3)),. rTh e empathy hierarchy here deals only with pronoun variables; we do not want two constant terms unifying via the hierarchy - which is doomed to failure.. 318. N(morph:X,pp:wa), C+(empathy(Zl,Z2,Z3)).. 3 T - s t r u c t u r e in D i s c o u r s e . 3 . 1 E m b e d d i n g a n d I n t e r l e a v i n g . In this section, we will illustrate some o f the ways in which T - s t r u c t u r e figures in J a p a n e s e discourse, s. W h a t we have below is a f a t h e r talking a b o u t t h e h e a l t h o f his children.. Chichioya<i> -wa 01<i> warat -te, father top laugh and. ~Taxo<h>-wa yoku kaze -wo hiku -n'desuyo. Taro top often cold acc catch aux-polite Kinou - m o 0 1 < t > kaze -wo hi'ire, 01<k> yesterday also cold acc catch gakko -wo yasu -n'da-n'desuyo. school acc take leave past aux-pollte Masako<j> -wa 01<./> gen'ldde, Ol<j> kaze. top healthy cold. -wo hi'ita koto -ga arimas en. acc caught experiende nora occur aux-neg 01<j> itsumo sotode ason'de -imasuyo.". often outdoors play aux-polite. - t o Ol<i> itta. comp said. "Taro often catches a cold. He got one yesterday again and didn't go to school. Masako stays in a good health and has n e v e r been sick with fin. I often see h e r playing outdoors." F a t h e r said with a smile on his face.. Here are t h e facts:(a) zero a n a p h o r a occurring within t h e q u o t a t i o n (internal a n a p h o r a ) are coreferential either with Taro or w i t h Masako; (b) those occurring outside (external a n a p h o r a ) , however, all refer t o chi- chioya; (c) chichioya has an a n a p h o r i c link which crosses over t h e entire q u o t a t i o n ; (d) syntactically, the q u o t e d p o r t i o n functions as a c o m p l e m e n t for the verb -to itta. It appears, moreover, t h a t an in- ternal a n a p h o r associates itself with Taro in case it occurs in the segment headed with Taro, an d with Masako in case it occurs in the segment headed with Masako. T h e n , since the q u o t e d discourse consists o f a set o f discourse segments, it will be assigned t o a T - s t r u c t u r e . B u t the s t r u c t u r e does not e x t e n d over t h e p a r t 01 itta, which completes the discourse, for t h e 01 c o r d e r s with chichioya, and neither with Taro or Masako. T h i s would give us an analysis like one in figure 4.. S Here and below we call a tree rooted at T a 'T- structure' and one rooted at D a 'D-structure'.. T. F i g u re 4: e m b e d d i n g . T h e following discourse shows t h a t t h e T - s t r u c t u r e can b e discontinuous:. [a] ~Masako<i> -ga kinou s i g o t o - w o nora yesterday work acc. yasun'da -n'desuyo." [b] Hahaoya<k> -wa took leave aux-polite mother nora. 01<h> isu -ni suwaru -to 0 1 < t > hanashi chair on sit when tell. hazimeta [c] "Kaze-demo 01<i> hi'ita -nolm." began, cold acc caught question. [ d ] - t o Chichioya-ga 03<k> tazuneta. comp father nom asked. "Masako took a leave f r o m the w o r k yester- d a y . ' , M o t h e r began to tell, as she sat on the chair. "Did she catch a cold f ", asked Father.. 01<i> c o r d e r s w i t h Masako, so [c] f o r m s a T - s t r u c t u r e w i t h [a]. B u t t h e two are s e p a r a t e d by a n a r r a t i v e [b]. Similarly, t h e coreference between 03<k> an d Hahaoya gives rise t o a T - s t r u c t u r e t h a t spans [d] an d [b], b u t t h ere is an i n t e r r u p t i o n b y nar- r a t i v e [c] (figure 5).. T T T . F i g u re 5: interleaving. 3 . 2 P r o b l e m . T h e r e is a curious i n t e r a c t i o n between a p a r a g r a p h - break an d a T / D - s t r u c t u r e . [Fujisawa et al., 1993], for instance, observes a s t r o n g t e n d e n c y t h a t J a p a n e s e zero a n a p h o r a are p a r a g r a p h - b o u n d e d . T h e following is f r o m N i h o n K e i z a i Shinbun, a J a p a n e s e economics newspaper.. K a w a m a t a H i d e o < i > . 01<i> Sagami tetsudo Mr. H. Kawamata Sagam/ Railways. kaichou. [San-gatsu] mik-ka gozen juichi-ji chairman March 3rd day a.m. 11-hour nijusan-pun, kokyuhuzen no-tame 23-mlnute respiratory insufficiency due-to. 319. Tokyo Machida de 01<i> sikyo, 01<i> nanajugo Tokyo Machida in dies 75 -Sai. yrs. old. T a n a k a Yutaka<k>. 01<k> Moto- Matsushita Mr. Y. Tanaka former Matsushita tsuushin kogyo senmu. [San-gatsu] telecom industries exective director March mik-ka gozen yo-ji san-pun, sin-huzen 3rd day a.m. 4-hour 3-mlnute cardiac failure no-tame Yokohama Midoriku de 01<k> sikyo, due-to Yokohama Midoriku in dies Ol<k> rokujuhas-sai.. 68 yrs. old. Mr. H. Kawamata, 75, chairman of Sagami-Railways, died of respiratory insuf- ficiency at 11:23 a.m., in Machida, Tokyo, March 3.. Mr. Y. Tanaka, 68, former executive direc- tor of Matsushila telecom industries, died of cardiac failure at 4:03 a.m., in Midoriku, Yokohama, March 3.. [Zero-anaphora are m a d e explicit here for expository purposes; they are not part o f the newspaper. The rest a p p e a r s as it does in the paper.] F r o m th e way s a m e - i n d e x a n a p h o r a are d i s t r i b u t e d over t h e dis- course, it seems r a t h e r obvious t h a t a p a r a g r a p h break has an effect o f m a r k i n g a segment for t h e discourse. 9 T h e present theory, however, fails t o deal with the s i t u a t i o n like this; it simply assigns a single DS s t r u c t u r e t o the discourse in question, giving a wrong i n t e r p r e t a t i o n t h a t zero a n a p h o r a present are all coreferential. As it stands, n o t h i n g in the t h e o r y provides for t r e a t i n g graphological m a r k s such as a p a r a g r a p h break. Yet, it is unclear to us wh et h er a p a r a g r a p h break is a signal for a I"- or D-structure.. 4 C o n c l u s i o n . We have developed a c o m p u t a t i o n a l t h e o r y for re- solving zero a n a p h o r a in Japanese, drawing on t h e results f r o m previous works on J a p a n e s e discourse [Kuno, 1987; K u n o and K a b u r a k i , 1977], etc). A m a- j o r d e p a r t u r e f r o m the t r a d i t i o n a l analyses o f zero a n a p h o r a lies in the reduction o f the space o f an- tecedents for zero a n a p h o r a . T h i s has been m a d e possible by a d o p t i n g ideas like Discourse Segment, Minimal Semantics Thesis and Empathy Hierarchy. In p a r t i c u l a r , we have shown t h a t the Minimal Se- m a n t i c s Thesis leads to reducing the n u m b e r o f an- tecedents for a segment to at m o s t three. Also shown in the p a p e r is t h a t the resolution o f zero a n a p h o r a is p a r t of parsing t e x t , so no additional m e c h a n i s m is. 9 We may note that a recursive embedding of discourse of the sort we have discussed above is effected through the explicit use of quotation marks; their absence would lead to the outright nngrammaticality.. needed. F u r t h e r m o r e , t h e present t h e o r y co m pares f a v o r a b l y with t h e previous schemes like Focusing an d Centering in t h a t it is able to deal with forward- an d backward-looking a n a p h o r a b y v i r t u e o f the way unification o p e r a t e s on t h e e m p a t h y hierarchy.. P a r t o f o u r discussion has t o u c h e d on t h e effect o f g r a p h o l o g y on t h e sem an t i cs o f discourse. T o date, no significant research has been done on t h a t area o f acad em i c interests. T h e l i t e r a t u r e suggest t h a t in t h e w r i t t e n language, t ex t s, i.e., cohesive discourses, are m a r k e d t h r o u g h a v ari et y o f linguistic an d non- linguistic means: n o n - a l p h a n u m e r i c ch aract ers (quo- t a t i o n m ark s, brackets, parentheses), g r a p h i c devices. t i n d e n t a t i o n , t a b u l a t i o n , i t e m i z a t i o n ) , an d so on Nunberg, 1990; Halliday an d I-Iassan, 1990]. T h u s a discourse segment m i g h t qualify for t h e t e x t h o o d since it has t h e p r o p e r t y t h a t zero p r o n o u n s are re- solved internally. Its i n d i c a t o r is, o f course, t h e topic particle wa. B u t for t h e T - s t r u c t u r e , it is far f r o m clear w h e t h e r it is a n y w a y cohesive, a n d i f it is, w h a t its i n d i cat o rs are. ( Q u o t a t i o n m a r k an d p a r a g r a p h b reak are possible c a n d i d a t e s . ) . S o m e o f t h e technical as well as linguistic details are y et t o be worked o u t ; we have n o t talked a b o u t how t h e topic comes t o b e associated w i t h one or m o r e zero p r o n o u n s in t h e segment. C o n s i d e r a t i o n s o n e m p a t h y m a y well influence t h e choice o f pro- n o u n s t o b e m a t c h e d with.. R e f e r e n c e s . [Fujisawa et al., 1993] Shinji Fujisawa, Shigeru Ma- s u y a m a , an d Shozo Naito. An In sp ect i o n o n Ef- fect o f Discourse C o n t r a i n t s p e r t a i n i n g t o Ellip- sis S u p p l e m e n t in J a p a n e s e Sentences. In Kouen- Ronbun-Shuu 3 (conference papers 3). I n f o r m a t i o n Processing Society o f J a p a n , 1993. In Jap an ese.. [Gazdar a n d Mellish, 1989] G e r a l d G a z d a r a n d Chris Mellish. Natural Language Processing in Prolog. Addison-Wesley P u b l i s h i n g Co., New York, 1989.. [Gunji, 1987] T a k a o G u n j i . Japanese Phrase Struc- ture Grammar. D. Reidel, D o r d r e c h t , 1987.. [Halliday an d Hassan, 1990] M. A. K. H al l i d ay a n d R. ttassan. Cohesion in English. L o n g m a n , New York, 1990.. [Hobbs, t o appear] J e r r y R. Hobbs. O n t h e Coher- ence an d S t r u c t u r e o f Discourse. in The Structure of Discourse, Livia Polanyi, ed i t o r, A b l ex Publish- ing Co., t o ap p ear.. [Inoue, 1978] K a z u k o Inoue. Nihongo -no Bunpo Kisoku ( Grammatical Rules in Japanese ). Taishukan, T o k y o , 1978. in J a p a n e s e . . [Kitagawa, 1982] C. K i t ag aw a. T o p i c c o n s t r u c t i o n in J a p a n e s e . Lingua, 57:175-214, 1982.. 320. [Kuno and Kaburaki, 1977] Susumu Kuno and Et- suko Kaburaki. Empathy and Syntax. Linguistic Inquiry, 8:627-672, 1977.. [Kuno, 1987] Susumu Kuno. Functional Syntax. The University of Chicago Press, Chicago, 1987.. [Kuroda, 1965] S. Y. Kuroda. Generative Semanti- cal Studies in the Japanese Language. Garland, New York, 1965.. [Longacre, 1979] R. E. Longaere. The paragraph as a grammatical unit. In Tamly Giv6n, editor, Syn- ta~ and Semancs vol. 1~. Academic Press, 1979.. [Mikami, 1960] Akira Mikami. Zon wa Hana ga Na- gai (The elephant has a long trunk.). Kuroshio Shuppan, Tokyo, 1960.. [Nomoto, 1992] Tadashi Nomoto. Discourse and se- mantics of zero-pronominals. In Proceedings of NLC workshop, Nagasaki, 1992.. [Nunberg, 1990] Geoffrey Nunberg. The Linguistics of Punctuation, volume 18 of CSLI Lecture notes. CSLI, 1990.. [Pereira and Warren, 1980] Fernando C. N. Pereira and David H. D. Warren. Definite clause grammar for language analysis - a survey of the formalism and a comparison with angumented transition net- works. Artificitial Intelligence, 13:231-278, 1980.. [Sidner, 1983] Candance L. Sidner. Focusing in the comprehension o f definite anaphora. In Brady and Berwick, editors, Computational Model of Discourse, pages 267-330. The MIT Press, Cam- bridge, 1983.. [Walker et al., 1990] M. Walker, M. Iida, and S. Cote. Centering in Japanese. In Proceedings of COLING '90, 1990.. 321

New documents

sentences is: 3a: John, said Mary, was the nicest person at the.. 3b: John said Mary was the nicest person at

Original Article Protective effects of Kangfuxin liquid Periplaneta Americana extract on chronic atrophic gastritis in rats via anti-oxidative stress and inhibition of COX-2..

Encoding con- fusion networks outperforms encoding the best hypothesis of the automatic speech recognition in a neural system for dialog state tracking on the well-known second Dialog

Outline how you are going to come into compliance with Regulation 5: Individual assessment and care plan: The nursing home strives to ensure that residents have individual care plans

Real-time dual-frame contrast-enhanced ultrasound mode was selected in imaging; real-time dynamic image storage technology was uses in the whole process for whole recording; the images

The objective of this analytical study is twofold: to explore the functions of silence with duration of one second and above, towards informa- tion flow in a dyadic conversation

This section sets out the actions that must be taken by the provider or person in charge to ensure compliance with the Health Act 2007 Care and Welfare of Residents in Designated

Compliance with the Health Act 2007 Care and Welfare of Residents in Designated Centres for Older People Regulations 2013, Health Act 2007 Registration of Designated Centres for Older

1Department of Urology, Guangdong Provincial Work Injury Rehabilitation Hospital and Jinan University, Guangzhou 510440, China; 2Department of Urology, Guangzhou First Municipal

Concretely, we observe that the model trained on news domain pays sim- ilar amount of attention to summary-worthy con- tent i.e., words reused by human abstracts when tested on news and