Automated Speech Act Classification in Tutorial Dialogue

Speech act classification is the task of detecting speakers\u27 intentions in discourse. Speech acts are based on the language as action theory according to which when we say something we do something. Speech act classification has various application in natural language processing and dialogue-based intelligent systems. In this thesis, we propose machine learning models for speech act classification that account for both content of the current utterance and context (previous utterances) of dialogue and we present this work on two domains: human-human tutoring sessions and multi-party chat based intelligent tutoring systems. The proposed speech act classification models were trained and tested on chat utterances extracted from the tutoring sessions and based on the domain specific properties of the datasets were designed to work with hierarchical and granular speech act taxonomies