Title: Next Utterance Prediction Based on Two Kinds of Dialog Models
Author: Yoichi Yamashita and Riichiro Mizoguchi
Reference: Proc. of 3rd Conference on Speech Communication and Technology (Eurospeech '93), Berlin, pp.1161-1164.

Abstract: This paper describes a method of predicting user's next utterances in spoken dialog based on two kinds of dialog models, SR-plan and TPN. The SR-plan is a low level model of interactions composed of a stimulus and a response and the TPN is a high level model representing transitions of topics in dialogs. The dialog manager predicts next utterances and provides the language processing unit with templates of case frames which are associated with SR-plans and instantiated according to the preceding utterances and the topic information. An experiment shows that the utterance prediction improves the performance of the utterance recognition and drastically reduces the search space in terms of candidates in the input word lattice.

Ftp article (ps-file, 4 pages, 287245 bytes)