Empirical Analysis of Three Dimensions of Spoken Discourse: Segmentation, Coherence, and Linguistic Devices

摘要: A discourse consists not simply of a linear sequence utterances,1 but meaningful semantic or pragmatic relations among utterances. Each utterance either bears relation to preceding constitutes the onset new unit meaning action that subsequent utterances may add to. The need model between such units and linguistic features is almost universally acknowledged in literature on discourse. For example, previous work argues for an interdependence particular cue words phrases as anyway, their location relative text, (e.g., Hirschberg Litman 1993, Grosz Sidner 1986, Reichman 1985, Cohen 1984); distribution duration pauses multi-utterance 1992, Chafe 1980, Butterworth 1980); form anaphoric noun current hierarchical actions, focus attention 1977, 1979, Passonneau 1985). However, there are variety distinct proposals regarding how three dimensions of: 1. sequences semantically pragmatically related utterances, 2. they reflect, 3. lexico-grammatical prosodic features. We refer these respectively segmentation, coherence, devices.