Text segmentation and topic tracking on broadcast news via a hidden Markov model approach.

Continuing progress in the automatic transcriptionof broadcast speech via speech recognition has raisedthe possibility of applying information retrieval techniquesto the resulting (errorful) text. In this paperwe describe a general methodology based on HiddenMarkov Models and classical language modelingtechniques for automatically inferring story boundaries (segmentation) and for retrieving stories relatingto a specific topic (tracking). We will present in detailthe features and...