tailieunhanh - Báo cáo khoa học: "An Information-State Approach to Collaborative Reference"

We describe a dialogue system that works with its interlocutor to identify objects. Our contributions include a concise, modular architecture with reversible processes of understanding and generation, an information-state model of reference, and flexible links between semantics and collaborative problem solving. | An Information-State Approach to Collaborative Reference David DeVault1 Natalia Kariaeva2 Anubha Kothari2 Iris Oved3 and Matthew Stone1 1Computer Science 2Linguistics 3Philosophy and Center for Cognitive Science Rutgers University PiscataWay NJ 08845-8020 Abstract We describe a dialogue system that works with its interlocutor to identify objects. Our contributions include a concise modular architecture with reversible processes of understanding and generation an information-state model of reference and flexible links between semantics and collaborative problem solving. 1 Introduction People work together to make sure they understand one another. For example when identifying an object speakers are prepared to give many alternative descriptions and listeners not only show whether they understand each description but often help the speaker find one they do understand Clark and Wilkes-Gibbs 1986 . This natural collaboration is part of what makes human communication so robust to failure. We aim both to explain this ability and to reproduce it. In this paper we present a novel model of collaboration in referential linguistic communication and we describe and illustrate its implementation. As we argue in Section 2 our approach is unique in combining a concise abstraction of the dynamics of joint activity with a reversible grammar-driven model of referential language. In the new information-state model of reference we present in Section 3 interlocutors work together over multiple turns to associate an entity with an agreed set of concepts that characterize it. On our approach utterance planning and understanding involves reasoning about how domain-independent linguistic forms can be used in context to contribute to the task see Section 4. Our system reduces to four modules understanding update deliberation and generation together with some supporting infrastructure see Section 5. This design derives the efficiency and flexibility of .