--------
ABSTRACT
--------

Constraint-based Dialogue Modelling: Indirect Speech Acts

The aim of the full paper is twofold: (1) to describe a
constraint-based framework for modelling dialogue and (2) the
formulation of a simple theory within the aforementioned framework for
modelling task-oriented dialogue fragments, in particular, fragments
involving indirect speech acts.

The idea of dialogue as a game, i.e., an activity which involves
rule-governed moves from one state to another, has become a popular
idea in research on dialogue (as witnessed by, for instance, many of
the papers in the proceedings of the annual workshop on the formal
semantics and pragmatics of dialogue which started in 1997, see Benz
and Jaeger, 1997). Despite the presence of such a common theme, it is
often difficult to compare alternative theories of dialogue. This is
due to fact that such theories are based on implicit assumptions
and formulated with the use of widely divergent formalisms (e.g,
plan-based AI formalisms, dynamic predicate logic, situation theory,
constructive type theory, discourse representation theory, etc).

It seems unlikely that the current multiplicity of formalisms for
dialogue modelling will disappear in the near future. Nevertheless, I
will try to show that it is possible to set up a general framework for
dialogue modelling which respects this multiplicity. The benefits of
such a framework are that the implicit assumptions underlying theories
of dialogue can be made explicit: a theory is obtained by filling in
the parameters of the framework. The framework includes a formal definition
of the notion of a dialogue. Both updating dialogue states and
generating dialogue actions are treated as constraint satisfaction
problems. Furthermore, I show how the framework enables us to obtain a
systematic overview of methods for relating formal theories of dialogue
to empirical evidence concerning dialogues.

In order to show how the framework relates to concrete theories of
dialogue, a simple theory is formulated within the framework. The
theory will be geared specifically to the treatment of indirect speech
acts. The two most important problems associated with the occurrence of
indirect speech acts (ISAs) in conversation are: (1) how do
interlocutors find out that a particular utterance has an indirect
meaning; (2) why do interlocutors generate ISAs instead of directly
saying what they want. Levinson (1983) argues forcefully that although
an answer to (1) can be formulated by means of a combination of Speech
Act Theory (Searle, 1969) and the Gricean Cooperative Principle (Grice,
1967), this does not resolve question (2). Levinson shows how both (1)
and (2) can be dealt with in Conversation Analysis.

By means of partial formalizations of the aforementioned approaches to
ISAs in my framework, I will argue that (A) question (1) is not
properly addressed by combining speech act theory and the Cooperative
Principle and (B) a structuralist account of ISAs, as in Conversation
Analysis, is incomplete with respect to utterances which have
repercussions beyond the bounds of a conversation (e.g., conditional
ISAs such as `Can you tell John that I want to speak to him, if you see
him?'). My alternative theory is intended to solve these problems. The
theory makes use of a Hamblin style (Hamblin, 1970) notion of dialogue
state in terms of commitments which interlocutors assign to each other
and to themselves.


References

Benz, A. and G. Jaeger (1997). Proceedings of MunDial 97, CIS 97-106,
University of Munich.

Grice, H.P. (1967). Logic and Conversation, Unpublished MS. of the
William James Lectures, Harvard University.

Hamblin, C.L. (1970). Fallacies, London: Methuen and Co. Ltd.

Levinson, S.C. (1983). Pragmatics, Cambridge: Cambridge University
Press.

Searle, J.R. (1969). Speech Acts, Cambridge: Cambridge University
Press.