On the role of context and prosody in the interpretation of ‘okay’

Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, pages 800–807, Prague, Czech Republic, June 2007. c⃝2007 Association for Computational Linguistics On the role of context and prosody in the interpretation of ‘okay’ Agust´ın Gravano, Stefan Benus, H´ector Ch´avez, Julia Hirschberg, Lauren Wilcox Department of Computer Science Columbia University, New York, NY, USA {agus,sbenus,hrc2009,julia,lgw23}@cs.columbia.edu Abstract We examine the effect of contextual and acoustic cues in the disambiguation of three discourse-pragmatic functions of the word okay. Results of a perception study show that contextual cues are stronger predictors of discourse function than acoustic cues. However, acoustic features capturing the pitch excursion at the right edge of okay fea- ture prominently in disambiguation, whether other contextual cues are present or not. 1 Introduction CUE PHRASES (also known as DISCOURSE MARK- ERS) are linguistic expressions that can be used to convey explicit information about the structure of a discourse or to convey a semantic contribution (Grosz and Sidner, 1986; Reichman, 1985; Cohen, 1984). For example, the word okay can be used to convey a ‘satisfactory’ evaluation of some entity in the discourse (the movie was okay); as a backchan- nel in a dialogue to indicate that one interlocutor is still attending to another; to convey acknowledg- ment or agreement; or, in its ‘cue’ use, to start or fin- ish a discourse segment (Jefferson, 1972; Schegloff and Sacks, 1973; Kowtko, 1997; Ward and Tsuka- hara, 2000). A major question is how speakers indi- cate and listeners interpret such variation in mean- ing. From a practical perspective, understanding how speakers and listeners disambiguate cue phrases is important to spoken dialogue systems, so that sys- tems can convey potentially ambiguous terms with their intended meaning and can interpret user input correctly. There is considerable evidence that the different uses of individual cue phrases can be distinguished by variation in the prosody with which they are re- alized. For example, (Hirschberg and Litman, 1993) found that cue phrases in general could be disam- biguated between their ‘semantic’ and their ‘dis- course marker’ uses in terms of the type of pitch accent borne by the cue phrase, the position of the phrase in the intonational phrase, and the amount of additional information in the phrase. Despite the frequence of the word okay in natural dialogues, relatively little attention has been paid to the rela- tionship between its use and its prosodic realization. (Hockey, 1993) did find that okay differs in terms of the pitch contour speakers use in uttering it, suggest- ing that a final rising pitch contour “categorically marks a turn change,” while a downstepped falling pitch contour usually indicates a discourse segment boundary. However, it is not clear which, if any, of the prosodic differences identified in this study are actually used by listeners in interpreting these po- tentially ambiguous items. In this study, we address the question of how hear- ers disambiguate the interpretation of okay. Our goal is to identify the acoustic, prosodic and phonetic fea- tures of okay tokens for …

Content truncated to 500 words