Automatic Extraction of Cue Phrases for Cross-Corpus Dialogue Act Classification

In this paper, we present an investigation into the use of cue phrases as a basis for dialogue act classification. We define what we mean by cue phrases, and describe how we extract them from a manually labelled corpus of dialogue. We describe one method of evaluating the usefulness of such cue phrases, by applying them directly as a classifier to unseen utterances. Once we have extracted cue phrases from one corpus, we determine if these phrases are general in nature, by applying them directly as a classification mechanism to a different corpus to that from which they were extracted. Finally, we experiment with increasingly restrictive methods for selecting cue phrases, and demonstrate that there are a small number of core cue phrases that are useful for dialogue act classification. 1 Motivation In this paper we present a recent investigation into the role of linguistic cues in dialogue act (DA) classification. Dialogue acts (Bunt, 1994) are annotations over segments of dialogue that...

• Type: Journal
• Year: 2010
• Where: COLING
• Authors: Nick Webb, Michael Ferguson
• Keywords: COLING 2010, Computational Linguistics, Cue Phrases, Dialogue, Dialogue Act