Management and Evaluation of Interactive Dialog in the Air Travel Domain

Introduction This paper presents the Unisys Spoken Language System, as applied to the Air Travel Planning (ATIS) domain. 1 This domain provides a rich source of interactive dialog, and has been chosen as a common application task for the development and evaluation of spoken language understanding systems. The Unisys approach to developing a spoken language system combines SUMMIT (the MIT speech recognition system [6]), PUNDIT (the Unisys language understanding system [3]) and an Ingres database of air travel information for eleven cities and nine airports (the ATIS database). Access to the database is mediated via a general knowledgebase/database interface (the Intelligent Database Server [4]). To date, we have concentrated on the language understanding and database interface components. A Dialog Manager integrates the overall user-system interaction. The Dialog Manager accepts user requests in the form of strings of words and calls PUNDIT to interpret the input; it then calls the database indirectly, via the IDS database interface. An important function of the Dialog Manager is to maintain a record of the discourse context, so that the system can successfully process connected dialog. We first describe our architecture in more detail, then give a short discussion of dialog management, a topic we feel will be crucial to successful systems interacting with users via natural language. We conclude with the presentation and analysis of results from the ATIS common evaluation task and from data gathered at Unisys using our system. Unisys DefenSe Systems Center for Advanced Information Technology PO Box 517 Paoli, PA 19301