Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

A Study of Broadcast News Audio Stream Segmentation and Segment Clustering

Matthew Harris, Xavier Aubert, Reinhold Haeb-Umbach, Peter Beyerlein

Philips Research Laboratories, Aachen, Germany

In transcription of broadcast news, dividing the signal into homogeneous segments, and clustering to-gether similar segments is important. Decoding a complete broadcast news program in one chunk is technically dificult. Also, through creation of homogeneous clusters of segments, improvement from adaptation can be increased. Two systems of segmentation and clustering are compared. The best system used the BIC algorithm to produce long, homogeneous segments, and a nearest neighbour bottom-up agglomerative clustering algo-rithm to produce homogeneous clusters. Adaptation brought aword error rate (WER) improvement from 23:4% to 21:0% using the automatic segmentation and clustering, compared to an improvement from 21:8% to 20:0% using a handmade \correct" segmentation and clustering.

Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Harris, Matthew / Aubert, Xavier / Haeb-Umbach, Reinhold / Beyerlein, Peter (1999): "A study of broadcast news audio stream segmentation and segment clustering", In EUROSPEECH'99, 1027-1030.