9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Is a Speech Recognizer Useful for Characteristic Analysis of Classroom Lecture Speech?

Kenji Kobayashi, Mitsuhiro Somiya, Hiromitsu Nishizaki, Yoshihiro Sekiguchi

University of Yamanashi, Japan

This paper investigates whether a speech recognizer is useful for characteristic analysis of lecture speech or not. It is important to pay attention to how we speak in order to effectively convey the meaning of what we are trying to say to someone. We have examined how various acoustic features included in a lecture speech influence the understanding, and change the impression, of the speech, which is determined by performing tests involving trial subject. The research investigated the correlation acoustic features in speech and the impression rating of students who listened the classroom lecture speech. In this paper, especially, we try to use a speech recognizer which translated speech into transcriptions with acoustic likelihoods. Recognizer's outputs may have hidden potential for rating speech. As an opening the research, we clear the relationship between the recognition rate of the lecture speech and the impression of students (or acoustic features of the speech).

Full Paper

Bibliographic reference.  Kobayashi, Kenji / Somiya, Mitsuhiro / Nishizaki, Hiromitsu / Sekiguchi, Yoshihiro (2008): "Is a speech recognizer useful for characteristic analysis of classroom lecture speech?", In INTERSPEECH-2008, 1341-1344.