My research is based on the conversation analytic approach to discourse interaction which examines the achievement of social activity through the use of language. Utterance types (e.g. questions, answers, assessments, excuses, complaints, assertions, challenges, et al.) achieve their meaning based on their structural location within the talk rather than as abstract context free linguistic forms.

A question, for example, in one instance may be hearable as a sardonic assessment, in another context a simple inquiry, and yet in a third context a denial of a request for information. The reason why the conversation analytic method is most suited for developing simulated dialogue, where a human can speak with a computer possibly as freely as with another human, is because it is premised on the concept that the analysis of conversation should be based on naturally occurring data rather than hypotheticals developed in laboratories.

The 25 years of research by coversation analysts who have approached this domain with the understanding that talk is systematically and strongly organized, have produced findings showing the structural organization of discourse as an activity which inheres regularities that are as much a part of the social nature of man as social conventions. This gives order to social life and enables the achievement of interactional work. By virtue of uncovering the systematic and strongly organized nature of conversational activities, one is then able to design programs based on this systematic and highly organized quality of discourse organization. What appears in natural language dialogue as the code of rules used by conversationalists can then be formulated into algorithms enabling computers to engage in natural language understanding.