An effective application of contextual information using adjacency pairs and a discourse stack for speech-act classification

A speech-act is a linguistic action intended by a speaker. Speech-act classification is essential to the generation and understanding of utterances within any natural language dialogue system as the speech act of an utterance is closely tied to a user intention. Lexical information provides the most crucial clue for speech-act classification, and contextual information offers additional complementary clues. In this study, we concentrate on how to effectively utilize contextual information for speech-act classification. Our proposed model exploits adjacency pairs and a discourse stack to apply contextual information to speech-act classification. Experimental results show that the proposed model yields significant improvements in comparison with other speech-act classification models as well as a baseline model, which does not utilize contextual information.