Temporal Confusion Network for Speech-based Soccer Event Retrieval

This paper introduces temporal confusion network and its application for speech-based soccer event retrieval, where an event is remarked by the announcer's spoken words. A temporal confusion network is a confusion network in which each cluster is marked with temporal information. Since the purpose of soccer event retrieval is to retrieve only the interesting highlights - not the whole video clip, temporal information is crucial in tracking them. By expanding the indexing model from 1-best transcriptions to temporal confusion networks, recall rates for event retrieval can be improved. Experiments are conducted on the first round of World Cup 2010 and the Vietnamese AFF Suzuki-cup 2008 databases. In the best case, an average improvement of 7.1% recall rate is achieved.