Publications

Advanced search

Abstract

Vicent Tamarit, Carlos D. Martínez-Hinarejos. Estimating the Number of Segments of a Turn in Dialogue Systems. Proceedings of the 9th International Workshop on Pattern Recognition in Information Systems - PRIS 2009, 2009. Ana Fred (Editors). pp. 9-17. INSTICC Press.

An important part of a dialogue system is the correct labelling of turns with dialogue-related meaning. This meaning is usually represented by dialogue acts, which give the system semantic information about user intentions. This labelling is usually done in two steps, dividing the turn into segments, and classifying them into DAs. Some works have shown that the segmentation step can be improved by knowing the correct number of segments in the turn before the segmentation. We present an estimation of the probability of the number of segments in the turn. We propose and evaluate some features to estimate the probability of the number of segments based on the transcription of the turn. The experiments include the SwitchBoard and the Dihana corpus and show that this method estimates correctly the number of segments of the 72% and the 78% of the turns in the SwitchBoard corpus and the Dihana corpus respectively.