The Effects of Interaction on Spoken Discourse

Near-term spoken language systems will likely be limited in their interactive capabilities. To design them, we shall need to model how the presence or absence of speaker interaction influences spoken discourse patterns in different types of tasks. In this research, a comprehensive examination is provided of the discourse structure and performance efficiency of both interactive and noninteractive spontaneous speech in a seriated assembly task. More specifically, telephone dialogues and audiotape monologues are compared, which represent opposites in terms of the opportunity for confirmation feedback and clarification subdialognes. Keyboard communication patterns, upon which most natural language heuristics and algorithms have been based, also are contrasted with patterns observed in the two speech modalities. Finally, implications are discussed for the design of near-term limitedinteraction spoken language systems. I N T R O D U C T I O N Many basic issues need to be addresssed before technology will be able to leverage successfully from the natural advantages of speech. First, spoken interfaces will need to be structured to reflect the realities of speech instead of text. Historically, language norms have been based on written modalities, even though spoken and written communication differ in major ways (Chafe, 1982; Chapanis, Parrish, Ochsman, & Weeks, 1977). Furthermore, it has become clear that the algorithms and heuristics needed to design spoken language systems will be different from those required for keyboard system s (Cohen, 1984; Hindle, 1983; Oviatt & Cohen, 1988 ~: 1989; Ward, 1989). Among other things, speech understanding systems tend to have considerable difficulty with the indirection, confirmations and reaffirmations, nonword fillers, false starts and overall wordiness of human speech (van Katwijk, van Nes, Bunt, Muller & Leopold, 1979). To date, however, research has not yet provided accurate models of spoken language to serve as a basis for designing future spoken language systems. People experience speech as a very rapid, direct, and tightly interactive communication modality, one that is governed by an array of conversational rules and is rewarding in its social effectiveness. Although a full. y interactive exchange that includes confirmatory feedback and clarification subdialo~mes is the prototypical or netural form of speech, near-term spoken language systems are likely to provide only limited interactive capabilities. For example, lack of adequate confirmatory feedback, variable delays in interactive processing, and limited prosodic analysis all can be expected to constrain interactions with initial systems. Other speech technology, such as voice mail and automatic dictation devices (Gould, Conti & Hovanyecz, 1983; Jelinek, 1985), isdesigned specifically for noninteractive speech input. Therefore, to the extent that interactive and noninteractive spoken language differ, future SLSs may require tailoring to handle phenomena typical of noninteractive speech. That is, at least for the near term, the goal of designing SLSs based on models of fully interactive dialogne may be inappropriate. Instead, building accurate speech models for SLSs may depend on