Paper Info
Title
Recent improvements to IBM's speech recognition system for automatic transcription of broadcast news.
Abstract
We describe extensions and improvements to IBM's system for automatic transcription of broadcast news. The speech recognizer uses a total of 160 hours of acoustic training data, 80 hours more than for the system described in Chen et al. (1998). In addition to improvements obtained in 1997 we made a number of changes and algorithmic enhancements. Among these were changing the acoustic vocabulary, reducing the number of phonemes, insertion of short pauses, mixture models consisting of non-Gaussian components, pronunciation networks, factor analysis (FACILT) and Bayesian information criteria (BIC) applied to choosing the number of components in a Gaussian mixture model. The models were combined in a single system using NIST's script voting machine known as rover (Fiscus 1997).
Year
DOI
Venue
1999
10.1109/ICASSP.1999.758056
ICASSP
Keywords
Field
DocType
gaussian mixture model,acoustic vocabulary,algorithmic enhancement,speech recognition system,bayesian information criterion,acoustic training data,single system,broadcast news,automatic transcription,mixture model,recent improvement,factor analysis,telephony,broadcasting,bayesian information criteria,gaussian processes,speech recognition,mixture models,hidden markov models,bayesian methods,information analysis,voting
Speech enhancement,Pronunciation,Bayesian information criterion,IBM,Computer science,Gaussian process,Natural language processing,Artificial intelligence,Pattern recognition,Speech recognition,NIST,Hidden Markov model,Mixture model
Conference
ISSN
ISBN
Citations 
1520-6149
0-7803-5041-3
14
PageRank 
References 
Authors
2.07
9
6
Name
Order
Citations
PageRank
S. S. Chen1333.41
E. M. Eide2594.55
Mark J. F. Gales33905367.45
R. A. Gopinath441748.03
Dimitri Kanevsky547754.37
P. Olsen68510.62