Dialogue Act Classification: Experiments with the SCHISMA Corpus

摘要: We describe experiments on dialogue act classification using empirical data. The datasets used have been extracted from the annotated schisma Wizard of Oz corpus. The features used for the classification task include utterance features that were derived from the output of a part-of-speech tagger and context features in the form of dialogue act types of previous utterances. The present experiments are primarily aimed at analysing the significance of the features in determining the most plausible dialogue act type of a given utterance.