Transcription of broadcast news-some recent improvements to IBM´s LVCSR system

Author

Polymenakos, L. ; Olsen, P. ; Kanvesky, D. ; Gopinath, R.A. ; Gopalakrishnan, P.S. ; Chen, S.

Author_Institution

Dept. of Comput. Sci., IBM Thomas J. Watson Res. Center, Yorktown Heights, NY, USA

Volume

2

fYear

1998

fDate

12-15 May 1998

Firstpage

901

Abstract

This paper describes extensions and improvements to IBM´s large vocabulary continuous speech recognition (LVCSR) system for transcription of broadcast news. The recognizer uses an additional 35 hours of training data over the one used in the 1996 Hub4 evaluation. It includes a number of new features: optimal feature space for acoustic modeling (in training and/or testing), filler-word modeling, Bayesian information criterion (BIC) based segment clustering, an improved implementation of iterative MLLR and 4-gram language models. Results using the 1996 DARPA Hub4 evaluation data set are presented

Keywords

Bayes methods; acoustic signal processing; broadcasting; grammars; information theory; speech recognition; speech synthesis; 4-gram language models; Bayesian information criterion; DARPA Hub4 evaluation data set; IBM LVCSR system; acoustic modeling; broadcast news transcription; filler-word modeling; iterative MLLR; large vocabulary continuous speech recognition; optimal feature space; segment clustering; testing; training; training data; Acoustic testing; Bandwidth; Broadcasting; Maximum likelihood linear regression; Speech analysis; Speech enhancement; Speech recognition; Telephony; Training data; Vocabulary;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech and Signal Processing, 1998. Proceedings of the 1998 IEEE International Conference on

Conference_Location

Seattle, WA

ISSN

1520-6149

Print_ISBN

0-7803-4428-6

Type

conf

DOI

10.1109/ICASSP.1998.675411

Filename

675411