A Multi-Purpose Model of Conversational Context

Natural language dialogue systems require con-textual information for a variety of processing functions, including reference resolution, speech act recognition, and dialogue management. While much has been written about individual contextual problems, many of the proposed representations are mutually incompatible , unusable by the agents involved in a conversation , or both. We present a model of context that tries to reconcile in a general and systematic fashion the diierences between the discourse models used for reference resolution, conversation act recognition , and dialogue management in a system dealing with conversations. Starting from the technical solutions adopted in DRT, we show rst of all how to obtain a discourse model that while preserving DRT's basic ideas about refer-ential accessibility, includes information about the occurrence of speech acts and their relations. We show then how the information about speech acts can be used to formalize the basic ideas of Grosz and Sidner's model of discourse structure. Finally, we extend this model to incorporate an account of the grounding process. 1 Motivations A vast amount of context is needed to participate in and understand even the simplest conversations. This context includes a large store of (assumed) common knowledge between the participants: about language itself, about conventions of social interaction (e.g., that a question should be responded to with an answer), about the domain of discourse (e.g., that trains move along rail lines, not through the air), as well as more local information shared between just the participants themselves. The individual mental states of the agents are an important element of the context of conversation. And previous utterances in the conversation itself will be important for recognizing and performing appropriate future utterances. This paper is about models of context for systems that engage in naturally occurring conversations. The motivation is provided by work on the TRAINS project at the University of Rochester, one of whose aims is the development of a planning assistant able to engage with its user in spoken conversations in the domain of transportation by train Allen et al., 1995]. The TRAINS prototype must perform at least three kinds of linguistic activities that depend on a context: reference resolution, conversation action recognition and dialogue management. The models of context for reference resolution are typically developed with the aim of making available the antecedents of various the emphasis in the models of context proposed for action recognition and dialogue management has been …