Topic Segmentation : A First Stage to Dialog-Based Information Extraction.
Narjès Boufaden, Guy Lapalme, Yoshua Bengio · 2001
We study the problem of topic seg-mentation of manually transcribed speech in order to facilitate informa-tion extraction from dialogs. Our approach is based on a combina-tion of multi-source knowledge mod-eled by hidden Markov models. We experiment with different combina-tions of linguistic-level cues on di-alogs dealing with search and rescue missions. Results show the effective-ness of multi-source knowledge.