Acoustic Analysis and Transcription of Linguistic and Paralinguistic Features in Dialogue Speech

Abstract

Nonverbal features of spontaneous speech are important in human dialogue. And transcription of dialogue including the paralanguage is a useful technique for analysis of the dialogues. But, the paralinguistic features are transcribed by transcriber's subjective now. Some features are measurable as concrete acoustical features. We made measurement of the speech rate through spectrogram transformation of the wave envelope and automatic estimation of such features is helpful in transcribing the dialogues. In this paper, we examine an automatic method of speech rate estimation and studied consistency of descriptions between different transcribers.