Speaker Diarization: About whom the Speaker is Talking ?
Julie Mauclair, Sylvain Meignier, Yannick Estève · 2006
The automatic speaker diarization consists in splitting the signal into homogeneous segments and clustering them by speakers. However the speaker segments are specified with anonymous labels. This paper suggests a solution to identify those speakers by extracting their full names pronounced in French broadcast news. A semantic classification tree is automatically built on a training corpus and associate the full names detected in the transcription of a segment to this segment or to one of its neighbors. Then, a merging method permits to associate a full name to a speaker cluster instead of an anonymous label provided by the diarization. The experiments are carried out over French broadcast news records from the ESTER 2005 evaluation campaign. About 70% show duration is correctly processed for both development and evaluation corpora. On the evaluation corpus, 18.2% show duration is wrongly named and no decision is taken for 11.9% show duration