Modelling Grounding and Discourse Obligations Using Update Rules

This paper describes an implementation of some key aspects of a theory of dialogue processing whose main concerns are to provide models of GROUNDING and of the role of DISCOURSE OBLIGATIONS in an agent's deliberation processes. Our system uses the TrindiKit dialogue move engine toolkit, which assumes a model of dialogue in which a participan. t's knowledge is characterised in terms of INFORMATION STATES which are subject to various kinds of updating mechanisms. 1 I n t r o d u c t i o n In this paper we describe a preliminary implementation of a 'middle-level' dialogue management system. The key tasks of a dialogue manager are to update the representation of dialogue on the basis of processed input (generally, but not exclusively, language utterances), and to decide what (if anything) the system should do next. There is a wide range of opinions concerning how these tasks should be performed, and in particular, how the ongoing dialogue state should be represented: e.g., as something very specific to a particular domain, or according to some more general theory of (human or human inspired) dialogue processing. At one extreme, some systems represent only the (typically very rigid) transitions possible in a perceived dialogue for the given task, often using finite states in a transition network to represent the dialogue: examples of this are systems built using Nuance's DialogueBuilder or the CSLU's Rapid Application Prototyper. The other extreme is to build the dialogue processing theory on top of a full model of rational agency (e.g., (Bretier and Sadek, 1996)). The approach we take here lies in between these two extremes: we use rich representations of information states, but simpler, more dialogue-specific deliberation methods, rather than a deductive reasoner working on the basis of an axiomatic theory of rational agency. We show in this paper that the theory of information states we propose can, nevertheless, be used to give a characterisation of dialogue acts such as those proposed by the Discourse Resource Initiative precise enough to formalise the deliberation process of a dialogue manager in a completely declarative fashion. Our implementation is based on the approach to dialogue developed in (Traum, 1994; Poesio and Traum, 1997; Poesio and Traum, 1998; Traum et al., 1999). This theory, like other action-based theories of dialogue, views dialogue participation in terms of agents performing dialogue acts, the effects of which are to update the information state of the participants in a dialogue. However, our view of dialogue act effects is closer in some respects to that of (Allwood, 1976; Allwood, 1994) and (Singh, 1998) than to the belief and intention model of (Sadek, 1991; Grosz and Sidner, 1990; Cohen and Levesque, 1990). Particular emphasis is placed on the social commitments of the dialogue participants (obligations to act and commitments to propositions) without making explicit claims about the actual beliefs and intentions of the participants. Also, heavy emphasis is placed on how dialogue participants socially GROUND (Clark and Wilkes-Gibbs, 1986) the information expressed in dialogue: the information state assumed in this theory specifies which information is assumed to be already part of the common ground at a given point, and which part has been introduced, but not yet been established. The rest of this paper is structured as follows. The theory of dialogue underlying the implementation is described in more detail in Section 2. Section 3 describes the implementation itself. Section 4 shows how the system updates its information state while participating in a fairly simple dialogue. 2 T h e o r e t i c a l B a c k g r o u n d One basic assumption underlying this work is that it is useful to analyse dialogues by describing the relevant 'information' that is available to each participant. The notion of INFORMATION STATE (IS) is therefore employed in deciding what the next action should be, and the effects of utterances are described in terms of the changes they bring about in ISs. A particular instantiation of a dialogue manager, from this point of view, consists of a definition of the contents of ISs plus a description of the update processes which map from IS to IS. Updates are typically triggered by 'full' dialogue acts such as assertions or directives, 1 of course, but the theory allows parts of utterances, including individual words and even subparts of words, to be the trigger. The update rules for dialogue acts that we assume here are a simplified version of the formalisations proposed in (Poesio and Traum, 1998; Traum et al., 1999) (henceforth, PTT). The main aspects of PTT which have been implemented concern the way discourse obligations are handled and the manner in which dialogue participants interact to add information to the common ground. Obligations are essentially social in nature, and directly characterise spoken dialogue; a typical example of a discourse obligation concerns the relationship between questions and answers. Poesio and Traum follow (Traum and Allen, 1994) in suggesting that the utterance of a question imposes an obligation on the hearer to address the question (e.g., by providing an answer), irrespective of intentions. As for the process by which common ground is established, or GROUNDING (Clark and Schaefer, 1989; Traum, 1994), the assumption in PTT is that classical speech act theory is inherently too simplistic in that it ignores the fact that co-operative interaction is essential in discourse; thus, for instance, simply asserting something does not make it become mutually 'known' (part of the common ground). It is actually necessary for the hearer to provide some kind of acknowledgement that the assertion has been received, understood or not understood, accepted or rejected, and so on. Poesio and Traum view the public information state as including both material that has already been grounded, indicated by GND here, and material that hasn't been grounded yet. These components of the information state are updated when GROUNDING ACTS such as acknowledgement are performed. Each new contribution results in a new DISCOURSE UNIT (DU) being added to the information state (Traum, 1994) and recorded in a list of 'ungrounded discourse units' (UDUS); these DUs can then be subsequently grounded as the result, e.g., of (implicit or explicit) acknowledgements. 3 I m p l e m e n t i n g P T T In this section, we describe the details of the implementation. First, in Section 3.1, we describe the TrindiKit tool for building dialogue managers that we used to build our system. In Section 3.2, we describe the information states used in the implementation, an extension and simplification of the ideas from PTT discussed in the previous section. Then, in Section 3.3, we discuss how the information state is updated when dialogue acts are observed. Finally, 1We assume here the DRI classification of dialogue acts (Discourse Resource Initiative, 1997). / . ' ° " .. -,. ' " . . , I.lol'lP.lllit~ ~;l{lle (i$)