5th International Conference on Spoken Language Processing

Sydney, Australia
November 30 - December 4, 1998

A Syllable-Based Chinese Spoken Dialogue System for Telephone Directory Services Primarily Trained with A Corpus

Yen-Ju Yang, Lin-Shan Lee

Dept. of Computer Science and Information Engineering, National Taiwan University, Taipei, Taiwan

This paper presents a syllable-based Chinese spoken dialogue system for telephone directory services primarily trained with a corpus. It integrates automatic phrase extraction, robust phrase spotting, statistics-based semantic parsing by phrase-concept joint language model as well as concept-based dialogue model, and intention identification by probabilistic finite state network to form a speech intention estimator. By applying the proposed techniques, the concept sequence with the maximum a-posteriori (MAP) probability based on intra and inter sentence consideration conveyed in the user's speech sentence, i.e. the speaker's intention, can be identified. This approach is convenient to be trained by a given corpus and flexible to be ported to different dialogue tasks. Incorporate a mixed-initiative goal-oriented dialogue manager, we have successfully developed a dialogue system for telephone directory service. Very promising results have been obtained in on-line tests.

Full Paper

Bibliographic reference.  Yang, Yen-Ju / Lee, Lin-Shan (1998): "A syllable-based Chinese spoken dialogue system for telephone directory services primarily trained with a corpus", In ICSLP-1998, paper 0528.