Dysfluencies, Exclamations and Laughter in Dialogue

点击翻译按钮获取中文摘要

Although dysfluent speech is pervasive in spoken conversation,dysfluencies have received little attention within formal theories ofgrammar---they are widely perceived as meaning-free productionerrors. The majority of work on dysfluent language has come frompsycholinguistic models of speech production and comprehension andfrom structural approaches designed to improve performance in speechapplications. Over recent years much evidence has accumulated thatdysfluencies, far from being meaningless noise, contain much usefulinformation that guides language users's actions and evaluations oftheir interlocuters' states of mind. Moreover, they exhibit rule-likeregularities on all levels (including phonology, syntax, andsemantics.). In DUEL we aim to show how dysfluent speech across anumber of languages (including French, German, English, and Chinese)can be analyzed in a precise way on the basis of formal grammaticaltools, using this theory to guide the design of dialogue systems whichcan deal head on with dysfluent speech, exploiting the informationtherein rather than filtering it away. DUEL will also tackle anotherphenomenon that has not hitherto received attention from formalgrammarians, namely laughter. Empirical studies over recent years haveshown that these occur relatively frequently in conversation and donot typically involve `humorous' utterances. They play a significantsemantic role, e.g. in indicating an utterance is not to be takenseriously or in enabling a socially delicate utterance to be madewithout causing offence. Our aim is to develop precise analyses of howlaughter is integrated in the emergence of meaning, precise enough toenable dialogue systems that understand and respond to laughter to beimplemented. The tools developed in DUEL to analyze disfluency andlaughter will enable a variety of other dialogical phenomena that havebeen somewhat marginal to be analyzed, e,g, exclamations, tagquestions, and corrective particles such as `No'. Both theory andimplementation in DUEL will draw on carefully collected parallel datain French, German, and Chinese, as well as a small number ofexperimental studies. This will enable subtle cross-linguistic andcross-cultural differences to be described, as well as deepercommonalities to be hypothesized.