Discourse System Project.

摘要

There has been a significant amount of work in the last decade on the processing of discourse, be it for modeling two-person extended dialogues, story and text understanding, or extended question answering systems. While there has been important progress in areas such as the use of world knowledge in language interpretation, the use of plan-based models of language (eg. speech act planning), models of reference and focus, and in models of discourse structure itself, work in each area has not been related to work in the other areas. No one has determined how each individual processing technique can be combined with the others to form a fully functional discourse system. Some suggestions on the organization of discourse have arisen recently (Grosz and Sidner, 1986; Litman and Allen, 1987; Allen, 1987), and while showing promise, have not yet been explored in enough detail for an actual application. This report describes an architecture for discourse systems that allows the integration of many different processing modules. In addition, it describes an initial set of modules that have been implemented and tested using the architecture. (KR)