A novel approach for robust dialogue act detection in a spoken dialogue system is proposed. Shallow\r\nrepresentation named partial sentence trees are employed to represent automatic speech recognition outputs.\r\nParsing results of partial sentences can be decomposed into derivation rules, which turn out to be salient features\r\nfor dialogue act detection. Data-driven dialogue acts are learned via an unsupervised learning algorithm called\r\nspectral clustering, in a vector space whose axes correspond to derivation rules. The proposed method is evaluated\r\nin a Mandarin spoken dialogue system for tourist-information services. Combined with information obtained from\r\nthe automatic speech recognition module and from a Markov model on dialogue act sequence, the proposed\r\nmethod achieves a detection accuracy of 85.1%, which is significantly better than the baseline performance of\r\n62.3% using a na�¯ve Bayes classifier. Furthermore, the average number of turns per dialogue session also decreases\r\nsignificantly with the improved detection accuracy.
Loading....