Detecting politician speech in TV broadcast news shows

Politician speaker turn detection in TV Broadcast News shows is addressed in this paper. After a first role labeling pass of speaker turns among anchor, reporter and other, turns labeled as other are submitted to a politician speech detection process. The proposed approach combines acoustical and lexical cues as well as contextual information, and does not use any specific politician model (person-independent). Experiments on a set of 101 TV broadcast news shows show that the proposed approach, which relies on fully automatic processing, enables to detect politician speech with an equal error rate of 12.1%, which turns to a maximal F-measure of 70.3% due to the unbalanced distribution among politicians and non-politicians.