Structured Information States: Raising and Resolving Issues

Update semantics models the increase of information as the elimination of epistemic possibilities in an information state. (Stalnaker 1979, Veltman 1996) Partition semantics of questions models a question as an equivalence relation over the epistemic possibilities in an information state. Each alternative class corresponds to one of the exhaustive answers to the question. (Groenendijk & Stokhof 1984) The combination, an update semantics for questions (USQ) models a query as a change to the information-structure of an information state. (J ager 1996) Each information state is structured by one or more issues: salient questions in need of an answer. I suggest to treat each alternative class as an epistemic possibility itself. In other words, possibilities that di er in facts not occurring in any issue, are treated equivalently. I believe this is in line with the philosophy behind update semantics: \ To understand a proposition to know the content of a statement or a thought is to have the capacity to divide the relevant alternatives in the right way. To entertain a proposition, i.e. to ask whether the proposition is true, is to focus one's attention on certain possibilities, contrasting them with others."(Stalnaker 1984, p 4,5) Information may grow in two ways: by eliminating alternatives or by adding or re ning issues. By adding or re ning an issue both the domain of discourse may grow (indistinguishable objects become distinguishable, irrelevant objects become relevant) and the set of possible predicates to describe and refer to objects becomes larger and more detailed. This insight may help to solve issues in natural language semantics that have to do, in one way or another, with relevance. I believe that an utterance is relevant, when it (partially) resolves (one or more of) the current issue(s), or in other words, when it gives a (partial) answer to (one or more of) the currently most salient question(s).1 It is notoriously hard to derive or predict the issue at a given point in discourse. However, for a xed task and a xed domain, heuristics can be given to predict the most likely issue. One such task domain is provided in the schismasystem. schismais an automatic inquiry system for the theatre domain. By the very set-up of the system we may assume that the user will have external perlocutionary goals: to go to the theatre, get information about performances and possibly order tickets. 1Modelling relevance by (partial) answerhood is not new. I was in uenced by Ginzburg (1995) Kuppevelt (1995) and Gerbrandy (p.c). 2 Such goals may help to predict issues.2 For automatic inquiry systems like schisma, issues can be raised in a number of ways: (1) by the user explicitly asking a question. (2) by the user introducing a new topic. A new topic normally raises `curiosity questions' about the nature of the topic. (3) by the user requesting an action of the server. Each action has preconditions. The truth of these conditions will become an issue. (4) by the system acting cooperatively. Actions can be planned to meet user goals. The truth of the preconditions of such actions will become an issue. (5) by the user explicitly protesting and indicating a change of issue. Assuming an issue can be predicted, USQ may be of use in analyses of the following puzzling phenomena: domain restriction, semantics of focus, scalar implicature, nominal anaphora resolution and the semantics of attitude verbs. References Ginzburg, J. (1995), `Resolving questions, I', Linguistics and Philosophy 18, 459{527. Groenendijk, J. & Stokhof, M. (1984), Studies on the Semantics of Questions and the Pragmatics of Answers, PhD thesis, University of Amsterdam. J ager, G. (1996), Only updates on the semantics of the focus particle only., in `Proceedings of the 10th AmsterdamColloquium', ILLC, Amsterdam. Kuppevelt, J. v. (1995), `Discourse structure, topicality and questioning', Journal of Linguistics 31, 109{149. Stalnaker, R. (1979), Assertion, in P. Cole, ed., `Pragmatics', number 9 in `Syntax and Semantics', Academic Press, New York, pp. 315{32. Stalnaker, R. (1984), Inquiry, MIT Press. Veltman, F. (1996), `Defaults in update semantics', Journal of philosophical logic 25(3), 221{262. previous version in DYANA R2.5A, 1990. 2I agree with Ginzburg (1995) that user goals are necessary in any theory of answerhood. 3