Fourth European Conference on Speech Communication and Technology

Madrid, Spain
September 18-21, 1995

Effectiveness of Pause Information in the Content Word Detection of Spoken Dialogues

Kazuyuki Takagi (1), Shuichi Itahashi (2)

(1) The University of Electro-Communications, Japan
(2) University of Tsukuba, Japan

One of the difficulties in spontaneous speech processing lies in the pausal phenomena. This paper describes an effect of incorporating pausal information into the syntactic rules, and an advantage of using an "utterance unit" as a processing unit in spoken dialogue recognition. A speech recognizer was implemented with utterance unit grammar and pausal information as word bigram probability. Recognition experiments were conducted using various goaloriented spoken dialogues. The method employing utterance unit grammar and pausal information showed best performance: content word detection rate is 50.4% for spontaneous speech, which was 8.8% better than the method based on sentence recognition without the constraints by pause information.

Full Paper

Bibliographic reference.  Takagi, Kazuyuki / Itahashi, Shuichi (1995): "Effectiveness of pause information in the content word detection of spoken dialogues", In EUROSPEECH-1995, 19-22.