<<

'HOW "/0 RESTRICT 0F DIBCOURSE

Barbara Dunin-K@plic z Institute of Informatics University Of Warsaw P.O. Box 1210 00-901 Warszawa, POLAND

ABBT~CT second interpretation seems better motivated. "/he next step is to define syntactic We single out a class of prototypes i.e., discourse domain, denoted by S_, i.e., a set L) a class of constructions forcing the obligatory of discourse expressions referring to or obligatory noncoreference. An individuals (set of individuals). The mapping essential feature of prototypes is their assigning individuals to syntactic expressions undistinctiveness. In this sense they are the will be ca/led the function and most natural and efficient mearis of denoted by R. F'orma/ly, R : S D 2 ID. in discourse. The non-application of prototype should be Example well motivated. This leads to the rule of (DI) John and Peter admire winter. They restricted choice stating that whenever it is are often skiing together. possible the application of a prototype should be preferred. SDI" - {"John", "Peter", "winter", "they '''~ The rule of the restricted choice suggests the general outline of interpreting ambiguous ID 1 - ~John, Peter, winter~ sentences, strictly speaking, the method of ordering admissible interpretations: those which ("John") - {John} can be equivalently expressed by means of a prototype are less probable. In words, R ("Peter") - {Peterl the rule of the restricted choice can be regarded as some kind of mechanism ordering R ("they") ,, {John, Peter]" the hypotheses for computer/on. R ("winter") - {winter}

INTRODUCTION We say that discourse expressions x and y a.re coreferencial, what we denote by xCy, The crucial problem in if and only if they refer to the same set of is the appropriate transposition of all individuals. expressions occurring in it, into reality (see, for instance, the framework provided by Kemp Formally, in (Kamp, 1981)). Even preliminary analysis shows that one real object can be identified for each x,y ~ S u xCy iff R(x)= R(y) by various surface constructions. This forces the necessity of dividing surface expressions It is readily verified that C is an into classes denoting identical individuals. equivalence relation. Obviously each equivalence class of C contains coreferentia/ The above problem can formally be stated expressions. The set of equivalence classes as follows. To each discourse D we assign of C will be ca/led the reference sta~te of some reality which can be understood as discourse and denoted by RSD. a set ID of individuals (semantic discourse domain) together with a set of relations Example defined on ID. The semantic discourse domain can be interpreted twofold: (D2) John took a knife. 1.o as a set of real objects i.e., objects ( RSD 2 ) " "~ knife '~-. existing in actual world; John 5~ ~"a 2 ° as a set of mental objects i.e., objects (D3) John took a knife. He hurt himself. existing in user's . (RSD3) ~"JOhn:', "he", "himself'.~ Although the first interpretation is more natural, it leads to some ontological problems, knife"}. concerning the distinction of fictitious and non-fictitious entities. Since there is no such We can observe here that adding new distinction from linguistic perspective the utterance to the discour.~;e changes its

93 reference . In this sense RSD is a Let us consider few instances of dynamic notion Let us note also that the prototypes. Because the ideas presented here problem of solution can be regarded are implemented for the Polish language, the as defining the relation C for the whole notion of prototype will be illustrated with discourse. a number of Polish sentences. An elided subject specific for inflexional can Both the speaker, while constructin~ be observed here. It is clenoted by (~ a discourse, and the hearer, while eunalysing Because elided subject expresses some it, try to achieve the identity of RSD at each step of the discourse. We argue in this aspects of thematic continuity, its interpretation seems to be an important step paper that to accomplish this effect, the during discourse analysis. English speaker has at his disposal (at each moment) translations of presented examples preserve a more restricted set of linguistic their syntactic shape. Unfortunately they are constructions than it seems intuitively. Let sometimes incorrect as English sentences. us notice that expressions belonging to one equivalence class have various syntactic (1.) Piotr obudzit si~1 ~)3.. podszec£~ do okna, shapes at different steps of discourse. It ca/'* ~2 ot~vorzy~ je i ~)3 wyskoczy{. be shown that the syntactic form of expressions at particular moments is not Peter woke up, ~I came to the window, accidential, i.e., elements of indicated •2 opened it and ~)3 jumped out. equivalence class are not interchangeable. Expressions: Peter, #I, ~2, ~)3 are coreferentiaL Another P R O'I'O'I'YP E S interpretation is unadm,ssible, in (I) we deal with obligatory coreference of expressions Recent discourse theories provide several (denoted by a - -- b). levels of larlguage analysis: morphological, syntactic, semantic and sometimes pragmatic. Each of these levels determines a (2) ~I~2 podszeci% do okna, characteristic set of notions and mechanisms. It is assumed here that the analysis of 82", utterance on each levels of language should 4~4 ~skoczy~ yield complete obtainable by tools available on this level ~)1. Woke up, q)2 came to the window, Classical anaphor resolvers act on semantic level on discourse analysis. ~Are take q)3 opened it and #4 jumped out. the position that for inflexion al languages the coreference relation can be partially described on the syntactic level An essential feature of in (2), similarly as in (a) (co-ordinate this partial characteristic is defining the clauses) and in (3), (4) (subordinate coreference relation quite apart from Peal clauses) the only acceptable interpretation individuals, i.e. without specyfing the reference is explicitely showed. function. "Po fix some ideas let us consider an (3) Z~im Ca,~s~ea~2 zgasi~ ~wiat~o. utterance containing the noun phrases NPl, ..., NP . If there is no information regarding n Before 911 left, 912 turned the light off. coreference, all we can do is to assert that the coreference relation is included between the "minimal" relation, i.e., relation identified (4) ~].~_Z_gasi~ ~wiat~o, zanim....w~2 wyszedL by the unit equivalence classes NP~ , ..., NP and the mcLximal one, i.e. adm,Ring in 911 Turned the L~ght off, before ~2 left. one nclass all number-gender agreeable phrases. The next examples concern the obligatory We very seldom deal with such a situation noncoreference of expressions (denoted by in practice. Almost a/ways we can assign to an utterance a syntactic level information a+-b) stating obligatory- coreference or obliqatory noncoreference of some expressions. (5) Ona lubi ja~ The surface constructions carrying this kind of information with respect to pronouns She likes her. and zero pronouns (in the case of elided subject) will be called prototypes. Ln other words prototypes can be regarded as (6) (~ zapyta~ Piotra,'~czy Jan p6jdzie syntactic means forcing obligatory coreference or obligatory noncoreference between pronouns do teatru. or zero pronouns and other surface expressions. ¢ Asked Peter, whether John would go to the theatre.

94 (7) ¢ Usiad~ do stc~u, a .Tan naleuI: mu wina. interpretations with respect to their plausibility. From the set of possible interpretations of Sat at the table, und John poured a , those that can be equivalently expressed by means of prototype, should be him out some wine. regarded as less plausible. The justification of this choice is clear: if the speaker wanted The above examples pose the question of to point out such an he would how the class of prototypes should be singled interpretation, naturally achieve it by applying a prototype. out. This problem can be solved by specifying a collection of rules concerning In view of the obove we can formulate the obligatory coreference and obligatory the rule of restricted choice. It states that noncoreference. The exact format of these whenever it is possible the application of rules is beyond the of this paper. a prototype should be preferred. For inflexional languages they depend on the It is irrportant to notice that the rule of type of considered sentence, the sentence- restricted choice can be viewed from the -level functions of considered phrases perspective both of the speaker contructing and their syntactic shape. As a simple the discourse and the hearer modelling it. example of such a rule let us consider the The speaker should apply prototypes basic criterion of excluding coreference: whenever it is possible. The hearer should take this fact into consideration. If the object is expressed by means of a reflexive pronoum, then it is coreferential Let us try to interpret the concrete with the subject; in other cases the sentences with the help of the rule of referential identily of the subject and object restricted choice. is excluded. (8) Zanim ~)I wyszed~, .Tan zgasi~ ~wiat{o. This criterion can be applied both for deterrninig coreferents of objects - blocking Before ~I leftmasc, .Tohn tumedmasc the subject, and in tesf/n~ the possible antecedents of the subject - blocking the the light off. objects. This is exactly the case we have in (5). There are two interpretations here:

THE RULE OF' RESTRICTED CHOICE (9) Zanim zgasi wia o

A conclusive criterion of being a prototype results from analysing a given sentence it% Before ~I left, John turned the light off. isolation. If it is possible to assert or to exclude the referential identity of some l expressions of the sentence, indepedently of (1.0) Zanim ~I w3zszed2, Jan zgasi{ ~wiat~o. its then the sentence can be regarded as an instance of prototype. An essential feature of prototypes is that they are Before ~ll left, John turned the light off. completely indistinctive and in this sense they are the most proper tool for expressing a certain relationship in the utterance. This (~ denotes the reference to the context). strong relationship makes it possible to eliminate some interpretations, which in other But the first interpretation can be expressed cases should be regarded as probable too. by means of prototypes. If we agree that achieving unambiguity of discourse is the major goal both for the (Pl) Zanim ~I ~Aryszed{, @2 zgasit ~wiat~o. speaker and the hearer, then the non- -application of prototype, as the most natural Before ~1 left, ~2 turned the Light off. and efficient mea/qs of communication should be well motivated. When such a special reason (P2) Jan zanim wyszedt, ~1 zgasit gwiatto. is lacking, the speaker should apply a a prototype. Under this assumption the set of John before left, ~1 turned the light off. linguistic tools available to the speaker is restricted. According to the rule of restricted choice The notion of prototype can be natural/y the first interpretation should be regarded as applied on the syntactic level of discourse less probable (note that it can be expressed anadysis to limit the number of hypotheses for by prototype). Hence the second interpretation further consideration. 13ut it can also be useful should be chosen. on the hi~her levels to interpret ambiguous discourses. Strictly speakin~ the properties of prototype suggest the general outline of interpreting ambiguous sentences, more precisely a method of orderin~ possible

9S Another example is more complicated. NONMONOTONICITY OF THE RULE 05' lqESTRICTED CHOICE (11) Zanim o__.nnwyszed~, ~I zapy~ak ~eo, czy #2 p6jdzie do kina. Consider the following example: (15) I

Zanim on wyszed~, @~zapytakj~o, ( 1,7 ) Kiedy ~p.odszed~ do Piotra, by~A, on czy ~2 p6jdzie do kina. zden erwowany.

According to our definition the above sentence 2ecause the second interpretation can is an instance of a prototype. unambiguously be expressed by the prototypical c on struction: Excluding the coreference of pointed expressions decreases the numer of possible (3.8) Kiedy ~I podszed{ do Piotra, by~ ~)2 interpretations, but does not clear up all referential relationships in this ambiguous zden er%vowalny. sentence. Although there are no further syntactic premises to resolve this ambiguity When ~)I came near Peter Q2 was we can specify the less probable interpretation by appying the rule of restricted choice. If the nervous. speaker wanted to express the following according to the rule of restricted choice the sense: first interpretaf/on should be preferred. (12) Zanim X wyszeck%, X zapy%a.% go, The rule of resfx'icted choice is based on the assumpt/on that whenever it is possible czy @1. p6jdzie do kineu . people use unambiguous constructions. Although usually va/id this assumption cannot he should have used the following (structural) be regarded as genera@ truth. This meeuns that prototype: the rule of restricted choice enables one to jump to plausible but not ironclad conclusions. (13) Z~im ¢~1 ~apyt~ go, "l~pically, such conclusions can be invalidated when new information is available. In our czy ~2 p6jdzie do kina. example the preferred intezq0retation might he overturned when we extend our discourse as (In such a sequence of clauses in the follows: sentence the rule of obligatory coreference (1.9) I

That was the result of an earlier

quarrel with Peter.

96 The neoessity of changing the preferred ACKNOWLED GEMEN'P interpretation follcws from the fact that new information is available. The property of I would llke to thank Dr Witold Lukaszewicz drawing plausible but defeasible inferences for helping me to clarify ideas presented in characterizes non-monotonic reasoning. this paper. Various forms of this kind of reasoning are now being developed (see (~AI-84)). It is now widely recognized that discourse understanding regulres nonmonotonic mechanisms in many aspects. The rule of Dunin-K@plicz B. (1983) Towards better restricted choice is an example of such understanding of anaphora, in: Prec. of a nonmonotonic tool the ist ACL Conference, Piss, 139-144. Kamp H. (1981) A theory of truth and semantic representation, in: 5"ormal methods in the study of language, C ONCLUSIONS IV[athematisch Centrum, Amsterdam, 277-322. (1.) While constructing discourse the speaker wants the hearer to understand him AAAI (1.984) Non-monotonic reasoning correctly. Even if he uses ambiguous workshop, New York 0 constructions he intends to cemrr, unicate the unique interpretations, and not to create in heater's mind a set of a/l possible hypotheses. It follows that constructhag NLU systems, which ~enerate all admissible interpretations, contradicts common sense reasonln~. So the essential problem is to determine methods o/ choosing the most appropriate interpretation. If this plausible interpretation fails, it should be revised.

(2) Employing the rule of restricted choice assumes the existence of some mechanism which determines whether a given construction can be regarded as a prototype. 'l~bis can be achieved by specifyins a set of rules quali~j'ing the obUgatory coreference and noncoreference of referrins expressions. A partied set of such rules for the l::ollsh language has been presented in (Dunin-K~plicz, 1983).

97