A joint intention-based dialogue engine

In this paper we discuss some of the advantages of using a joint intention-based interpreter for building spoken dialogue systems. We describe STAPLE [1], a joint-intention interpreter that enables a system to obtain team and communicative behavior automatically without having to program this behavior explicitly. With this approach there is no necessity for a programmer to indicate when a question should be posed, when information should be shared etc. The interpreter enables the agents to exhibit team-oriented dialogue by interpreting the constructs of Joint Intention Theory(JIT) along with first principles reasoning over a formal semantics of communicative acts [2, 3]. We try to show that STAPLE can subsume and extend the finite-state and frame-based dialogue approaches available today from commercial dialogue systems. In particular, we show how STAPLE can handle over-answering, dynamic environment changes and teamwork in a general domain independent manner.