Machine interpretation of bilingual dialogue

This paper examines the role of the dialogue manager component of a machine interpreter. It is a report on one project to design the discourse module for such a voice-to-voice machine translation (MT) system known as the Interpreting Telephone. The theoretical discourse framework that underlies the proposed dialogue manager supports the job of extracting and collecting information from the context, and facilitating human-machine language interaction in a multi-user environment. Empirical support for the dialogue theory and the implementation described herein, comes from an observational study of one human interpreter engaged in a three-way, bilingual telephone conversation. We begin with a brief description of the interpreting telephone research endeavor, then examine the discourse requirements of such a language-processing system, and finally, report on the application of the discourse processing framework to this voice-to-voice machine translation task.