PARSING JAPANESE HONORIFICS IN UNIFICATION-BASED GRAMMAR

Hiroyuki MAEDA, Susumu KATO, Kiyoshi KOGUREand Hitoshi IIDA

ATR Interpreting Telephony Research Laboratories Twin 21 Bldg. MID Tower, 2-1-61 Shiromi, Higashi-ku, Osaka 540, Japan

Abstract expressed by special linguistic means independent of the This paper presents a unification-based approach to • utterance situation. For example, by violating a usage Japanese honorifics based on a version of HPSG (Head-driven principle for the determination of an honorific attitude (i.e. Phrase Structure Grammar)ll]121. Utterance parsing is based "one should not exalt oneself in front of others"), uses of an on lexical specifications of each lexical item, including honorific expression about the speaker himself can function honorifics, and a few general PSG rules using a parser capable as a kind of joke. However, without the help of grammatical of unifying cyclic feature structures. It is shown that the properties of honorifics independent of particular utterance possible word orders of Japanese honorific predicate situations, the violation of a usage principle itself could not constituents can be automatically deduced in the proposed be recognized at all, thus the expression could not function as framework without independently specifying them. a joke. Though the former situational determination of Discourse Information Change Rules (DICRs) that allow honorific attitude is an interesting subject matter for socio resolving a class of anaphors in honorific contexts are also and psycho-linguistic researchers, the latter grammatical formulated. properties of hot~orifics are our concern here and what is described with lexical specifications for honorifics. 1. Introduction 2.2. Mizutani's Theory of Honorific Expression Act Japanese has a rich grammaticalized system of honorifics to express the speaker's honorific attitudes toward discourse Mizutani's theory of honorific expression act is agents (i.e. persons who are related to the discourse). As introduced to define basic honorific attitude types that opposed to such written texts as scientific or newspaper stipulate the pragmatic constraints on Japanese honorifics. In articles, where the author's rather 'neutral' honorific attitude this model, discourse agents are positioned in an ~bstract is required, in spoken dialogues, an abundant number of two-dimenslonal honorific space (Fig. 1). How they are honorific expressions is used and plays an important role in positioned is a socio and psycho-linguistic problem, which is resolving human zero-anaphors. not pursued here. In this paper, a unification-based approach to Japanese honorifics is proposed. First, Mizutani's theory of honorific expression actl3] is introduced to define basic honorific Agent P (px,py) attitude types used in specifying pragmatic constraints on the Hearer (hx,hy) e e~,~ use of Japanese honorifics. Then a range of honorifics are classified into subtypes from a morphological and syntactico- Speaker (0,0) ~" semantic perspective and examples of their lexical specifications are shown. The main characteristics of the Agent Q (qx,qy) I utterance parser and an approach to explaining possible word orders of honorific predicate constituents are described. Fig 1. Honorific Space Finally, Discourse Information Change Rules are formulated An honorific expresson act reflects the configuraion of that resolve a class of anaphors in honorific contexts. these discourse agent points. The speaker is set as the point of origin, and the speaker's honorific attitude toward a 2. Speaker's Honorific Attitudes toward Discourse discourse agent, say P, is defined as the position vector of Agents point P. The speaker's honorific attitude toward agent P 2.1. Grammatical Aspects of Honorifics relative to agent Q is defined as a vector from point Q to A distinction must be made between the speaker's point P. The value and the direction of the vector are defined honorific attitude as determined by the utterance situation as follows: (the social relationship between discourse agents, the atmosphere of the setting, etc), and the honorific attitude as

139 Honorific Value : While regular-form honorific words share a common base for v = (x. y), the honorific value of a vector v (written with their nonhonorific forms because they are derived by as IvJ) is defined as: the productive honorific-affixation process, irregular-form Ivl = y iffx=0; honorific words have special word forms that have no direct 0 iffx ~0; connection to their nonhonorific forms. This distiction plays Honorific Direction : an important role in the lexical specification of honorifics and a. up I,t>0, in possible word orders of Japanese honorific predicate b. down Ivi < O, constituents. c. flat Iv~=O and x=O, d. across Ivl = 0 and x ~ O. 3.1.2. Syntactico-Semantic Viewpoint IN.B.J Assumingan honorific space to be two dimensional (not one In traditional school grammar, Japanese honorifics have dimensional),an across direction can be distinguishedfrom a fiat direction. An acrosSdirection of a vector correspondsto the case where no positive been classified into three categories: respect words honorific relation between the two agents (i.e. up, down, or flat) is ('sonkeigo'), condescending words ('kenjougo'), and polite recognizedby the speaker. words ('teineigo'). However, in this traditional tripartite classification, common features of respect-words and Though the speaker's honorific attitudes can be condescending-words not shared by polite-words are not characterized from several viewpoints (e.g. up/down, explicit. That is, while an agent toward whom the speaker's distant/close, formal/informal), Mizutani's model is honorific attitude is expressed must be grammatically located appropriate for describing Japanese honorifics because the in the sentence (i.e. as subject or object) in the case of respect up~down aspect most relevantly characterizes Japanese or condescending words, this requirement does not apply to honorifics. Moreover, it is not clear how the other aspects are polite words. Thus a more elaborate classification is adopted. independently grammaticalized in the Japanese honorific Conventional terms are replaced by Haradal4l's more system. syntactico-semantically motivated ones. Based on the direction of the vector defined above, the following four subtypes of honorific attitude relations are Classification by the syntactic role of an aqent to whom the distinguished. speaker's honorific attitude is expressed: honorific-word = Honorific Attitude Type : a. propositional-honorific-word= a. honor-up a. 1. subject-honorific-word (respect-word) b. honor-down (e.g. "kudaser-u" [give~nf]) c. honor-flat a.2.object-honorific-word(condescendiog-word) e. honor-across (e.g. "sesiage-ru" lgivev~ef]) b. performative-honorific-word (polite-word) (e.g. 'des-u', 'mas-u') 3. Description of Japanese Honorifics IN.B.] For example,a verb which takes a nonanimetesubject (e.g. "fur-u" 3.1. Classification of Japanese Honorifics in the sentece "Ame (rain) ga (SBJ)fur-u(fall). ° IThe rain falb.]) can be put 3.1.1. Morphological Viewpoint into its performative honorific form ('Ame ga fur-i mas-u.'), but not into its subject honorific form (* "Ante ga o-for-t ni nar-u.'). This is in accordance In Japanese, words in a wide range of syntactic categories with the difference between propositional honorifics and performative (i.e. nouns, verbs, adjectives, nominal-verbs, nominal- honorificl. adjectives, etc) are systematically put into their honorific IN.B.] Thereare a classof words which function in between the a.2 and b forms. They are classified into two subtypes according to how types of honorifics (e.g. "mair-u" [go/come~] in "Basu ga mair-i mas-u." they are derived from their nonhonorific forms. [A bus will come.]). Let us call them propositional-performattve-wordl.

Classification by the lexical derivation type: Minus-honorifics are given no place in the traditional honorific-word = tripartite classification. However, they are classified in our a. regular-form-honorific-word approach as correponding to the expressed honorific attitude (e.g. "ookak-i" from "kak-i" [writevinf]) types. [HP-[writevstem-CSinf]l b. irregular-form-honorific-word (e.g. "ossyar-" from "iw-" [speakvstem]) |N.B.] HP and CS stand for 'Honoric Prefix' and 'Conjugation Suffix' respectively. Wordsis transcribedin its phonemicrepresentation.

140 Classification by the expressed honorific attitude type: [[orth(orthography) ?orth] honorific-word = [head [[pos(part-or-speech) v] a. plus-honorific-word [ctype(conJugat|on-type) vowel] [cform(conJugatton-rorm) stem]I] (e.g. "aw-a-re-ru" [meetregular.sbjhon]) [adjacent ?prod] [{.,.meet~tem'CSvong]-PlusHonAuxv~tem-CSaml] [subcat ( b. minus-honorific-word ?sbJ[[haad [[pos p] (e.g. "aw-i-yagar-u" [meetregular-sbjhon]) [grf(grmmaticel-functton) sbJ]]] [subcat 0) [[...meet~em-C$1nf|-MinusHonAux~tm-CSm~J [sam ?sbJsem] IN B.] The Japanese honorific system has no systematized means to [sear [[huNn +]]]] positively express honor-flat or honor-across honorific attitudes. An non- ?prad[[heed [[pos v] [ctype ?predctype] honorific plain word form may expresshonor-flat honorific attitudes towerd [cforll vong(vofce-nagattva)] a discourse agent in a situation such as speaking to an old friend, while it may [subcet {~sbJ}] express honor-across honorific attitudes in a situation such as writing a [sea ?predsem]])] technical paper. [Sam ?predsam] [prsg [[restrs {[[reln honor-up] [origin espeakar e] Because the classfications of honorifics from different [goal TshJsem]])]]]]) viewpoints as summarized above are cross-categorical, and where E (<'ra" cons> thus independent of one another, a single honorific word <'rBre" (:or vowel kuru (e.g. "hozak-u" [sayvsenf]) can function at the same time as suru)>) irregular-form-honorific-word, subject-honorific-word, and Fig 2. Lexical Specification for a simple subject-plus minus-honorific-word. honorification morpheme ('(ra)re-ru') IN.g,] ? ~ a prefix for a tag-name used to represent a token identity of 3.2. A Unification-based Lexical Approach feature-~ru~ures. *Speaker* is a special global variable bound to a feature A unification-based lexicalism approach is adopted here stru~ure representing the speaker's information. for describing Japanese honorifics for the following reasons: The 'prag' feature describes the pragmatic constraint on (a) a unification-based approach enables the integrated description of information from various kinds of sources this expression (the "honor-up" relationship from the (syntax, semantics, etc), thus allowing their simultaneous speaker to the subject agent of the predicate is required for this expression to be used in a pragmatically appropriate analysis; way). Description with the 'honor-up' honorific attitude (b) a lexical approach helps to increase the modularity of grammar. In this approach, a grammar has only a small relation shows that this expression is a 'plus-honorific' expression. Structure-sharing of the 'goal' feature value of number of general syntactic rule schemata and most of this honorifc attitude relation with the semantic value of the grammatical information is to be specified in a lexicon. Linguistic word-class generalizations can be formed by predicate's subject shows that this expression is a 'subject- honorific' expression. The requirement for the 'orth' feature making grammatical categories complex by representing value (?'orth) and the 'ctype' value in the 'subcat' feature them with feature-structures. (?predctype) describes the morphophonemic characteristic of The specification of verbal category honorifics is this morpheme by stipulating that 're-(ru)' subcategorize for important because the verbal categories are the most either a regular consonant-stem ctype verb or an irregular productive in the honorification process, and thus ctype verb ('suru'[do]), and that 'rare-(ru)' subcategorize for appropriate to clearly show how diverse aspects of the either a regular vowel-stem ctype verb or an irregular ctype Japanese honorific system are described in this approach. verb ('kuru' [come]), correctly allowing (la) and (lc) but not 3.3. Examples of lexical specifications fib).

3.3.1. Regular-Form Honorifics (1) a. Sensei ga kyoositu e ika re to. Subject Honorification by "Vvong + (ra)re-ru" teacher $8J classroom to golctYoe vowell Past "(The) teacher went to (the) classromm.° Regular form honorifics are compositionally analyzed by b. *Sensei ga kyoositu e ika rar.__ee ta. giving lexical specifications for each honorific-word formation formative. For example, most plain-form verbs can c. $ensei ga kyoositu e ko rare to. comelctvoe kuru] be put into their simple subject-plus-honorific form by "(The) teacher came to (the) classroom." postpositioning the auxiliary verb "(ra)re-ru" to them ('re- d. *Kyoositu • ko $ensei ga rare to. ru" and "rare-ru" are allomorphs of a single morpheme). Lexical information for these formatives is specified in the feature structure:

141 The 'adjacent' feature is a special feature which assures adverbial postpositions appearing in other environments that its value be the first element in the list when the set (e.g. "Sensei ga kyoositu DAKE e WAko rare ta" [the teacher description in the 'subcat' value is expanded into list came only to the classroom] ); descriptions by a rule reader. The specification of this feature (c) a naive analysis fails to explain the kind of the elements implies that this morph is a bound morph and thus requires its that can operate as a Vinf element in the pattern, which is adjacent element to be realized as a nonnull phonetic form. automatically explained in the proposed framework as will be Though the set description in the 'subcat' value is introduced shown in section 5. to allow word order variation among complement daughters This regular object-plus-honorification process is in Japanese, without this kind of specification, compositionally analyzed in the proposed framework by ungrammatical sequences such as (ld) are also allowed for giving each of its formatives a lexical specification, inthe same auxiliariy verbs. manner as the "Vvong + (ra)re-ru" pattern subject-plus- [N.B.) A set description in the subcat feature of a feature honorific analysis. sturucture,[ladjacent ?c][subcatETa ?b ?c)]|, for example, is expanded into its Here the expression "o-aw-i-suru" is analized. Fig 3.a corresponding two possible list descriptions by a rule reader as follows: represents the lexical information of the verb "aw-' (meet) in I[adjacent 7c)[subcat (:or <7c ?b ?a> )]. Furthermore, its infinitive form ('aw-i'). . for example, is expanded into a feature structure such as [Jfirst ?c][rest[Ifirst 7bnrest Ilfirst ?a][restend]]. [[orth "aw-t"] (cen-tsKe-hp +][lex ~] [head [[pos v](ctype cons][cform tnf] Object Honorification by "HP + Vinf + suru" [hpforll "O']]] Next, let us consider a more complicated formation (subcet [[(heed [(pos p][grf sbJ][rom gel (seer ([hullan +]]]]] pattern for deriving a regular object-plus-honorific form. As [subcat {)] productive as the above "Vvong + (ra)re-ru" pattern is, an (sell ?sbJsell]] [[head [[pos p][grf obJ][fom nf] "HP +Vinf +suru" pattern can put most verbs with two (sellf ((hullan +]]]]] grammatical human arguments into their corresponding (subcat ()) [see ?ohJsmel]))) object honorific forms as follows: (sell [[reln meet] [agent ?sbJsell] [object ?obJsee]]]] "o + aw-i + suru" from "aw-" (meetvstem), Fig 3.a. Lexicallnformationfor "aw-i" (meetvinf) "go + shoukai + suru" from "shoukai" (introduce-verse). IN.B.] "o," and "go-" are two forms of s single morpheme (honorific prefix) that is prefixed to words in a variety of syntactic categories (See First, honorific prefixation lexical rule is applied to this Appendix I). The choice depends on the following element's origin. If the infinitive-form verb. Fig 3.b represents the lexical element is a Sine-Japanesemorpheme (kango), the honorifc prefix takes the information of an honorific prefix (HP) and Fig 3.c shows how form "go-'; if it is a native one, the honorific prefix is realized as "o-', this lexical rule is stated in the proposed framework. though there are exceptions. [[o rth ?hpform] In a naive analysis of Japanese honorifics, these honorific [head [(pos hp] (coh ([can-take-hp +][lex +] forms derive from their corresponding plain forms by a simple [head ([pos v][cforlx fnf] object honorification lexical rule that does not take into [hpform 7hpfom])]]]]] [subcet 0)) account their internal constituent structures (e.g. "aw-u" --) Fig 3.b. Lexicalinformation for HP preceding Vinf "o-aw-i-suru'). Accordingly, this kind of naive analysis is inadequate for the following reasons: (defrule x -> (hp x) Ca) it is arguable that "HP+Vinf" forms a unit in some (C0 can-take-hp) -- -) structural level before forming the unit "HP + Vinf + suru', ((1 head coh> -- (2)) ((0 head> -= C2 head>) considering the existence of such constructions as (C0 subcut) -- C2 subcut)) "lIP+Vinf+ni+nar-u" (normal-sbj-plus-hon-form), ((0 sell> -- C2 Sell>) ((0 pro 9 rsstrs) -- (:union C! prog restrs) "HP + Vinf + negaw-u(request)', and "HP + Vinf + itadak- (2 preg restrs)))) u(receive-favorirregular.obj.plus.hon.form)', but this assertion is Fig 3.C. Honorific prefixation rule not explicitly illustrated in a naive ana4~sis; IN.B.i The rule stated in an extended version of PATR41 notation consists (b) though some adverbial postpositions such as "we" of two parts; CFG-part and constraints. CFG-part is used to propose an (contrastive), "me" (also) and "sae" (even) can appear inside efficient top-down expectation in the parser. Constraints are required for the object honorific form (e.g. "o-aw-i-WA-suru', "go- the rule application to end successfully. Here, all constraintsare describedby equations of two feature structures. °< >" is used to denote a feature shoukai-SAE-MO-suru'), it is difficult to derive these forms by structure path, and ° ,," to denote a token identity relation between two a naive analysis in light of the generalization concerning feature structures.

142 The 'headlcoh(CategoryOfHead)' feature of a category 3.3.2. Irregular Form Honorifics specifies the kind of its head. An HP can take a lexical Irregular form honorifics share most of their lexical infinitive-form verb whose 'can-take-hp' value is' + '. An HP is information with their nonhonorific counterparts. In our assigned its appropriate realization form (.) (in this case, "o" framework, redundant lexical specification for irregular-form form), because its 'orth' value and the head's 'hpform' value honorifics is avoided by using lexical inheritance mechanism are the same. The first equation in the rule statement from their superclassas. For example, the necessary lexical prevents a second application of the honorific prefixation specification for the irregular subject honorific form "(- rule to the same verb (*'o-o-aw-i °) by specifying that the te)itadak-" of the donatory auxiliary verb "(-te)moraw-" is mother category's 'can-take-hp' feature value be ,., (**) The reduced, as shown in Fig 4.a. This turns out to be equivalent other equations in the rule are ones common to the adjunct- to Fig 4.b by unifying pieces of information from its super- head structures. classes, te-receive-favor and obj-plus-hon. I*N.B.] A note is needed here concerningthe realization of Hr. When the adjacent feature of the second right-hand-side symbol in the CFG-part is nil (:supere]asses to-receive-favor obJ-p]us-hon) [[orth "ftadak"] as in the above case, it is enough just to concatenate both 'orth' feature [head [[ctypa cons][cform stem]]]]) valuesof the right-hand-sidesymbols and make it the 'orth' feature value of the left-hand-side symbol. However. when the head element's adjacent Fig 4.a. Neccesarylexical specification for the irregular form feature has a nonnull value (i.e. in the casethat the head element is n bound donatoryauxiliaryverb'~te)itadak-" morph)o a more complicted operation is needed. But here we only mention [[orth "ftedak ° ] its necessityand avoid its preciseformulation to savespace. [head [[pos v][ctype cons][cform stem]J] I**N.BJ The 'can-take-hp' feature is specified as '-' not only for already HP- [subcet {[[head [[pus p][grf sbJ][form g8]]] prefixed elements, but also for almost all irregular form honorific verbs (e.g. [zuhcat {}] *'o-osshar-i'lsay], *'o-itadak-i'lreceive*favorD and most mono-synablic [sam ?sbJsem]] infinitive-form verbs that have correspondingirregular-form honorifics (e.g. [[head [[pus p][grf obJJ[fons nt]]] [subcJt {}] *'o-si" [doJ, *'o-mi" [look atJ). [sam ?ob~sem]] [[head [[pus v][cform teJ]] [subcat {[[heed [[pus p][grf sbJ]]] Next, the usual complement-head structure rule (Fig 3.d) [subcat (}] is applied to the resulting feature structure for "o-aw-i" and [see 7obJsa=]]}J [sam ?predsem]]}] the feature structure for a normal object-plus honorification [Sell [[reln transfer-favor] formative ('-suru', as shown in Fig 3.e). Thus the normal [donator ?zbJsam] [donatea ?ob~sem] object plus honorifc form ('o-aw-i-(suru)') for "aw-'[meet] is [accmepenfed-actton ?predsem]]] obtained in a compositional way. [prag [[rostra {[[reln honor-up] [orfgfn ?sbJsem] [go81 ?obJsam]J (derrule m -> (c h) [reln empathy-degree] ((0 heed> -- <2 head>) [more ?sbJsem] (<1> -- (:ftrst <2 subcat>) [lass ?ohJsemJ]J]]]]) ((0 subcat> -- (:rest <2 subcat>)) ((0 sam> -- <2 sam>) (<0 prag restrs> (:union (1 prag restrs> Fig 4.b. Whole lexical Information for "(-te)itadak-" (2 prag restrs>))) Fig 3.d. Complement head structure rule Lexical Information for other irregular-form honorifics is likewise specified. [[orth "'] [heed [[pus v][ctype suru][cform stem] [frregular-crorms [[vong sf][inf sf]''']]]] 4. Unification-based CFG Parser [can-take-hp -] Fig 5 shows the organization of the unification-based CFG [adjacent ?prod] [subcat (?sbJ[[head [[pos p][grf sbJ] parser. The parser is essentially based on Earley's algorithm, [samf [[human +]]]]] and unifies feature structures in its completion process. The [subcat (}] [sem ?sbJsem]] description of grammatical rules and lexical items are ?obJ[[hend [[pos p][grf obJ] complied into feature structures by the rule reader. [semr [[hu.en +]]]]] [subcat {}] Unification of cyclic feature structuers might be necessary [sam ?ohJsem]] to analyze certain expressions. To give some examples: ?prod[[head [[pos v][cform tnf][hp +]]] [subcet {?sbJ ?obJ}] (a) frozen honorific words such as "o-naka" (belly) and "go- [scm ?prsdsem]]}] ran" (to look at) must always be prefixed by an HP (the [sam ?predsem] [prag [[restrs {[[reln honor-up] element in bold face); [or4gtn ?sbJsem] (b) the polite form ('gozar-') of the verb "ar-'/'ir-" (to be) [gee] ?obJsem]]}]]]] almost always needs to be followed by the polite honorific Fig 3.e. Lexical Specification for a normal object-plus honorification formative ('(-suru)') auxiliary verb "-masu" in modern Japanese.

143 explanations for them were given from a rather speculative ~'~ SauceWmww perspective. In this research, it is shown how possible word orders can be deduced from lexical specifications of I ~"~" I t honorifics.

5.1. Propositional and Performative Honorifics UtterancePmrser based on Earley'salgorithm I A propositional honorific formative always precedes a performative honorific formative. For example, though ~l~ ~-~Festwestm(t~emtlficJitl(m ] "awa-re-masu" ([[[meetvong]-SbjPIusHon]-PerformativeHon]) I and "o-awi-si-masu" ([[[HP-meetvlnf]-ObjPlusHonJ- PerformativeHon]) are possible expressions, they would be impossible if their word orders were reversed (i.e. performative honorific placed before propositional Fig 5. Organization of the Unification-based Parser honorific). This restriction on word order is considered a In describing the above linguistic phenemena, it is convenient consequence of the lexical specifications for both types of if requirements f.or its head category can be specified not only honorifics. As shown in section 3, propositional for adjunct elements, but also for complement elements. In honorification formatives subcategorize a verbal category such cases, one more equation as follows needs to be added whose subject (and object) elements are not filled yet as its to the usual head-complement structure rule statement adjacent element. On the other hand, a performative shown in Fig 3.d. honorification formative subcategorizes a verbal category <1 headcoh> .. <2> with saturated subcategorization. This represents the lexical specification for "mesu °. The complied feature structure for the equations in Fig 3.d [[orth "'] plus the above equation includes a cyclic structure as shown [heed [[pos v][ctype musu][cforll stem] in Fig 6 [4rrugullr-cforlu [[senf mesu]...]]]] [cen-tlko-hp -] An extended version of WroblewskilS]'s feature structure [adjacent ?prod] unification algorithm was developed to allow rule statements [subcut {?prud[[heud [[pos v][cform musu]]] [suhcet (}] including cyclesl61. The extended algorithm can unify cyclic [sea ?predsum]]]J feature structures while avoiding unnecessary overcopying of [sims ?prudsms] [prig [[restrs {[[reln honor-up] feature stuructures. [ordgdn Ospuakure] loom1 *hem.re]]}]]]] 5. Word Order of Honorific Predicate Constituents Fig 7. Lexical Specification for a performative honorification In Japanese, a verbal predicate is composed of one main formative "masu" verb and postpositioned auxiliary verbs (though possibly none exist). Because both main verbs and auxiliary verbs may The performative honorificaton formative "masu" have honorific forms, various sequences of honorifics might cannot, therefore, immediately precede a propositional be expected to occur in a predicate as a simple matter of honorification formative due to the requirement concerning possible combinations. However, their possible word orders the adjacent element of propositional honorifics. The are restricted by a grammatical principles. Traditionally, opposite order, however, constitutes a syntactically possibile word orders were described in detail and the legitimate structure.

5.2. Subject and Object Honorifics An object honorific formative must precede a subject honorific formative, though there is an important class of exceptions (verbs that subcategorize a 'te' form verb as an adjacent element such as "(-te)itadak-'[receive-favor]). For s REST example, "o-awi-sa-reru" ([[[HP-meetvtnf]-ObjPlusHon|- SbjPIusHon]) is a possible word order, but "o-awa-re-suru" ([[HP-[meetvong-SbjPlusHon]]-ObjPlusHon]) is not possible if "-re(ru)" is used as an honorification formative. This word Fig 6. Cyclic part of the compiled feature structure order restriction can be explained in the same way as for the

144 above case: that is, as shown in section 3, the normal object [[sam [[cent ?xOl[[reln 'meet] [agent espeaker*] honorification formative %suru" subcategorizes a verb [object ?xO2] whose subject and object are not yet filled. The simple subject [t;oc ?xO3]]] honor|float|on formative "-(ra)reru" that requires its object to [fnds { ?xO4[[ver ?xO2[[type 'fnd]]] be already filled cannot, therefore, precede the normal [fem41tartty '-] [restrs (?x0S[[reln 'computettonal- subject plus honorification formative on account of lfngu4st] conflicting specifications for the 'subcat' value. Otherwise, [fnstance ?xO2]] ?xO6[[reln 'famous] no conflict exist. [Instance ?x0Z]3)]333. Other kinds of restrictions on the possible word order of ?x07[[var ?x03[[type 'tloc]]] [famtlfartty '-] Japanese honorific predicate constituents can likewise be [restrs [?xOa[[reln "hold| explained in the proposed framework. [object ?xO9] [tloc ?x03]] ?xlO[[reln "temporally-precedes| 6. Anaphora Resolution in Honorific Contexts [ante ?x03] [post "no.']]}]]]] In Japanese honorific contexts, many human anaphors ?xll[[ver ?xOg[[type 'fnd]]] can be resolved by recourse to pragmatic constraints on the [fam411artty '-] [restrs {?x|Z[[reln 'namtng] use of honorifics. This is an attempt to apply DR theory to the [name 'a01-88] anaphora resolution in Japanse honorific contexts. [namod ?x0033}333333 [prag [[restrs [affect the theory. Let k be a current DP,S, o be a linguisticstructure for an input utterance unified from lexicalspecifications, and k' be a DRS to be obtained. [[unfv ([[rm espeakare[[type 'tndfvtdual]]] DICR 1. (i) if o~sem~contis typed as a "non-quantified- [[l'm eheeureC[type 'tnd4vtdual]]] circumstance',then [[rm *now*[[type 'temporal-location||| Jim *heree[[type 'spatfo1-1ocatfon]]]}] kluniv - kluniv U oisem[indslvar,and [conds {}3] klconds - klcondsU oJsemlcontU otsemlindsJrestrs. (ii) if olsemlcontis typed as a 'universally-quantified- Fig 8.a. Initial DRS circumstance',then (N.B.1} Referencemarkers for the indexicals are directly anchored to kluniv - k[univ,and objects in the world, but the anchoringinformation is not shown here. kJ~onds - k[condsU {[(reln ',e|lante kl]lpost k2]]} where k I and k2 are newly introducedORS$ whose Now let (3a) represent a discourse-initial utterance. information contentsare specifiedbemcl on the o~Lsemlcontlquantlindvalue and the dsem[contlscope value as follows (3) a. Izen ACL-88 ga hiraka-re ta toki, watasi wa aru chomei-na keisan-gengogaku-sha ni o-a| si masi ta. "Once when ACL-88 was held. I met (object-honorific and DICR 2. kluniv . kJuniv,and performative-honorific)a certain famouscomputational linguist. ° k'lconds - kjcondsU dpraglrestrs Fig 9.a. Discourse Information Change Rules (part) From this, Fig 8.b is unified as its semantic/pragmatic information. The method of specifying necessary lexical For o to be felicitousw.r.t, k, it is requiredfor everyindex i in o that: (i) if i~familiarity - ' -, then i[variablef kJuniverse. information was briefly explained in section 3. (ii) if i[familiartty - ' +,then (a) ilvariable( kluniverse,and The initial discouse information is updated by the (b) ilrestriction is unifiable with kJcondition. semantic/pragmatic information of a new utterance as Fig 9.b. Novelty Familiarity Condition follows: First, DICR 1, shown in Fig 9.a below, is applied to the semantic value of a new utterance. DICR 2 is then applied index in the semicontJinds value has a Ifamiliarity'-] attribute in to the pragmatic value. Meanwhile, anaphoric expressions in Fig 8.b. a new utterance are resolved so that the NFCIS| shown in Fig ([[unfv [[[rm espeaker.]] [[rat ehearer.]] 9.b below is observed. [[rm *now*]] [[rm *harem|| [[rat ?x02]] [[m ?x033] [[m ?x0033}3 [conds (?x0! ?xg5 ?x06 ?x08 ?x|0 ?x;2 ?x13 ?x14]]]] In this case, Fig 8.c is obtained as an updated DRS, because the Fig 9.b. Updated DRS type of semlcont value is a 'basic-circumstance' and every

145 In this context, assume (3b) is uttered, Fig 8.c is its unified (3) b'. Song keisan-gengogaku-sha wa watasi ni aisatu nasal semlprag values. masi ta. "That computational linguist greeted (subject-honorific and (3) b. ?Sono keisan.gengogaku-sha wa watasi ni aisatu si perfor mative-honoriflc) me." yagari masi ta. "That computational linguist greeted (subject-minus-honorific and IN.L) Our DICRI with NFC also explain the failure of coindexing "song performative-honorific) me." keisan-gengogaku-she" in (4b) with a universally quantified expression °done ... me" (every ...) in a previous utterance, because the reference markers introduced for a universallyquantified expressionare in sul:mrdiate [[sam ]]cent ?xlS[[reln 'greet] [agent ?xl6] DRSs by OICR 1 end not accessiblefrom "song keisan-gangogaku-she"as a [recipient *speaker*] possibleantecedent. ) [tloc 7x17 ]]] [tnds (?x18[Cvar ?xlG[(typa 'lnd)]] (4) e. Izen ALL-88 ni sanka sl ta toki, watad via done charnel.ha kelsan- [familiarity '+] gengogeku.sha rd me o-el si meg ta. [restrs { "When I once took part in ACL-88, I met (object-honorific and ?xlg[[raln "computational- performative-honorific) every famouscomputational linguist." linguist) ]Instance ?xl6]])]] b. ? Song keisan-oenoooaku-sha we watasYniaisatunasaimesita.($b~ ?20[[var ?17[[typa 'tloc]] [restrs { ?21[[raln ' tlmpor811y- Though many issues rermain unaddressed concerning precedes] [ante 717] anaphora resolution in Japanese honorific contexts, these can [post *noo']))])]]] be approached by use of the proposed model. This model [prag [[restrs (?22[[roln 'honor-down) [agent *speaker*] regards discourse understanding as the process of unifying [object (16)]] various kinds of partial information, including contextual ?23['[reln 'honor-up) [agent *speakere] information. [object *hearer*]])]]]]]] Fig 8.c. Resulting Semantic Information for (3b) 7. Condusion A unification-based approach to Japanese honorifics Because the index 7x18 for "song keisan-gengogaku-sha" based on a version of HPSG was proposed. Utterance parsing (that computational linguist) has a ]familiarity '+] attribute is based on the lexical specifications of a range of honorifics based on the lexical specification for 'song', an attempt is using a parser capable of unifying cyclic feature structures. made to resolve it by unifying 7x16 with an element of the The developed parser constitutes an important part of kluniv value, requiring that their restrictions can also be NADINE (NAtural Dialogue INterpretation Expert), an unified. It stands to reason that it can be resolved experimental system which translates Japanese-English because 7x16 and 7x02 are, semantically speaking, unifiable, telephone and inter-keyboard dialogues. because their semantic restrictions are {[]rein 'computational- linguist]!instance 7x16]]} and [[[reln 'computational- linguist]linstance ?x02]]Ilreln 'famous)[instance ?x02]]) respectively, and Acknowledement their variable types are both 'individual', which causes no The authors are deeply grateful to Dr. Kurematsu, the president of ATR Interpreting Telephony Research Laboratories. Dr. Aizawa, the head of incompatibility. However, their pragmatic restrictions Linguistic Processing Department, end all the members of Linguistic ({llreln 'honor-downJlagent %peeker*)lob]act 7x16|] [[reln "honor- ProcessingDepartment for their constanthelp end encouragement. upJlagent %peaker*]lobject "hearer*]]}, and {([reln 'honor-up)[agent References *speaker*)lob]act ?x02]] ]It*In 'honor-up]iagent *speeker*]lobject [1] Pollard,Carl & Ivan Sag, 1967, Information-BasedSyntax and Semantics. *hearer*)l}) prevent ?x16 from being unified with ?x02, due to vol. 1. CSLILecture Notes 13. the stipulation 'llreln 'honor-up][agent ?ailobject ?b]] A [Ireln 'honor- 12] Genii,Takao. 1987.Japanese Phrase Structure Grammar. Reidel. [3] Mizutani. Sizuo., 1963, "Taiguu Hyougen no Sikumi." (Structure of down)[agent ?el]object ?b)] - bottom'. This anaphoric resolution Honorific Expressions),in Unyou(The Progmatics).Asakura. therefore fails. Other ways of resolving this anaphoric [4] Harada.S. I., 1976, "Honorifics." in Shibatani(ed.), Syntaxand Semantics 5. AcademicPress. expression also fail because of the incompatibility of their IS] Wroblewski, David A., 198, "Nondestructivegraph unification." in the variable types or semantic features. In any case, utterance sixth conf. on AI. [6] Kogure, Kiycsi, et al. 1988 (forthcoming), "A Method of Analyzing (3b) turns out to be infelicitous by NFC. Japanese Speech Act Types." in the 2nd conf. on Theoretical and Unlike (3b), utterance (3b'), whose sem/prag values are MethodologicalIssues in MachineTranslation of Natural Languages. [7] Kemp,Hans., 1981, "A Theoryof Truth and SemanticRepresentation." in the same as Fig 8.c except for [[rein 'honor-up)[agent Groenendijk et el. (ads.), Formal Methods in the Study of Language. *speaker*)lob]act ?x16]] instead of []rein 'honor-down)[agent Mathametisch Centrum. 18] Helm. Irene. 1963, "File Change Semanticsand the Familiarity Theory of *speaker*)]object ?x16]], can be given a felicitous reading, Definiteness." in BSuerle et al. (ads.), Meaning, Use and Interpretation of because anaphora resolution is possible without violating Language. Waiter de Gruyter. NFC in this case,

146