2018/05/16 by Chandrakant Bothe, Cornelius Weber, Bothe, Chandrakant +5
Computer Science · #Speech and dialogue systems #Topic Modeling #Natural Language Processing Techniques
paper · pdf · doi:10.48550/arxiv.1805.06280
Dialogue act recognition is an important part of natural language\nunderstanding. We investigate the way dialogue act corpora are annotated and\nthe learning approaches used so far. We find that the dialogue act is\ncontext-sensitive within the conversation for most of the classes.\nNevertheless, previous models of dialogue act classification work on the\nutterance-level and only very few consider context. We propose a novel\ncontext-based learning method to classify dialogue acts using a character-level\nlanguage model utterance representation, and we notice significant improvement.\nWe evaluate this method on the Switchboard Dialogue Act corpus, and our results\nshow that the consideration of the preceding utterances as a context of the\ncurrent utterance improves dialogue act detection.\n