Differences in the Speaking Styles of a Japanese Male According to Interlocutor; Showing the Effects of Affect in Conversational Speech

Computational Linguistics and Chinese Language Processing Vol. 12, No. 1, March 2007, pp. 1-16 1  The Association for Computational Linguistics and Chinese Language Processing [Received September 30, 2006; Revised October 30, 2006; Accepted November 1, 2006] Differences in the Speaking Styles of a Japanese Male According to Interlocutor; Showing the Effects of Affect in Conversational Speech Nick Campbell∗+ Abstract There has been considerable interest recently in the processing of affect in spoken interactions. This paper presents an analysis of some conversational speech corpus data showing that the four prosodic characteristics, duration, pitch, power, and voicing all vary significantly according to both interlocutor differences and differences in familiarity over a fixed period of time with the same interlocutor. Keywords: Conversational Speech Corpus, Expression of Affect, Prosodic Characteristics, Voice Quality Analysis. 1. Introduction Human spoken interactions convey a variety of different types of information. In addition to the linguistic content of speech, there are also paralinguistic and extralinguistic elements that convey discourse-level and interpersonal levels of information related to the speaker, to the speaker’s relationship(s) with the listener, and to the intended and actual progress of the discourse [Lindblom 1990; Stenström 1994; Hirschberg 1992, 1995]. Affect is conveyed in speech communication in a multitude of ways [Cahn 1989], including facial expression, gesture, body posture, speaking-style, tone-of-voice, lexical choice, syntactic construction, etc. It is perhaps impossible for a human to speak without revealing information about his or her affective states [Campbell 2005]. This paper examines how such affective information might be carried in the voice, particularly in the prosody of the speech, and shows from an examination of some corpus data ∗ National Institute of Information and Communications Technology + Spoken Language Communication Research Laboratory, Advanced Telecommunications Research Institute International, Keihanna Science City, Kyoto 619-0288, Japan. E-mail: nick@nict.go.jp; nick@atr.jp 2 Nick Campbell that evidence can be found for changes in affective state according to the nature of the interocutor and the history of their discoursal relationship. 2. The JST/CREST ESP Corpus Speech research nowadays is predominantly corpus-based. One learns about the characteristics of speech and the expressivity of speech utterances from the analysis of a very large number of samples collected under a variety of speaking conditions [Campbell et al. 2006; Cowie et al. 2005]. For a period of five years, in order to aid the development of a technology capable of Expressive Speech Processing (ESP), the Japan Science and Technology Agency funded the collection of a large corpus of expressive speech that was coordinated by ATR in Kyoto, Japan. As part of this corpus, over a period of three months during 2002, a group of ten volunteers were employed to talk with each other over the telephone for half-an-hour each time and to record their conversations to DAT using high-quality head-mounted condenser microphones. These conversations and their manually-produced transcriptions now form subset ESP_C of the JST/CREST ESP Corpus. Of the ten volunteers, two were native speakers of Chinese, both fluent in Japanese, one male and one female, and two were native …

Content truncated to 500 words