Corpus-based discourse understanding in spoken dialogue systems

摘要: This paper concerns the discourse understanding process in spoken dialogue systems. enables system to understand user utterances based on context of a dialogue. Since multiple candidates for result can be obtained utterance due ambiguity speech understanding, it is not appropriate decide single after each utterance. By holding results and resolving as progresses, accuracy improved. proposes method this statistical information from corpora. Unlike conventional methods that use hand-crafted rules, proposed easy design process. Experiment have shown exploits performs sufficiently effective.