Why machines should analyse intention in natural language dialogue

Abstract

One of the most difficult problems in Artificial Intelligence (AI) is to construct a natural language processing system which can interact with users through a natural language dialogue. The problem is difficult because there are so many ways by which a user can phrase his/her utterances to such a system. An added problem is that different types of users have different types of intentions and will conduct different exchanges with the system. While many have proposed theories and models of the processing of intentions in dialogue, few of these have been incorporated within working systems and tested empirically. Here, an experiment is conducted to test what we call the Intention-Computer Hypothesis: that the analysis of intention in natural-language dialogue facilitates effective natural-language dialogue between different types of people and a computer. The experiment provides evidence to support the hypothesis. In turn, the hypothesis provides evidence for a theory of intention analysis for natural-language dialogue processing. A central principle of the theory is that coherence of natural-language dialogue can be modelled by analysing sequences of intention. A computational model, called Operating System CONsultant (OSCON), implemented in Quintus Prolog, makes use of the theory and hypothesis to understand, and answer in English, English questions about computer operating system.