Abstract | ||
---|---|---|
In this paper we report on the various techniques that we implemented in order to improve the basic speech recognition performance of the BYBLOS system. Some of these methods are new, while others are not. We present methods that improved performance as well as those that did not. The methods include Linear Discriminant Analysis, Supervised Vector Quantization, Shared Mixture VQ. Deleted Estimation of Context Weights, MMI Estimation Using "N-Best" Alternatives, Cross-Word Triphone Models. While we have not yet combined all of the methods in one system, the overall word recognition error rate on the May 1988 test set using the Word-Pair grammar has decreased from 3.4% to 1.7%. |
Year | DOI | Venue |
---|---|---|
1989 | 10.3115/1075434.1075475 | HLT |
Keywords | Field | DocType |
mmi estimation,linear discriminant analysis,context weights,overall word recognition error,improved performance,basic speech recognition performance,high performance speech recognition,deleted estimation,shared mixture vq,byblos system,improved hmm model,cross-word triphone models,markov processes,vector analysis,error rate,speech recognition,word recognition,quantization,linear systems,discriminate analysis | Triphone,Pattern recognition,Computer science,Word error rate,Word recognition,Speech recognition,Vector quantization,Artificial intelligence,Linear discriminant analysis,Hidden Markov model,Quantization (signal processing),Test set | Conference |
ISBN | Citations | PageRank |
1-55860-112-0 | 4 | 4.40 |
References | Authors | |
5 | 11 |
Name | Order | Citations | PageRank |
---|---|---|---|
S. Austin | 1 | 138 | 71.34 |
Chris Barry | 2 | 20 | 9.70 |
Yen-Lu Chow | 3 | 148 | 70.09 |
Alan Derr | 4 | 18 | 14.79 |
Owen Kimball | 5 | 83 | 17.82 |
Francis Kubala | 6 | 410 | 99.88 |
J. Makhoul | 7 | 1097 | 233.37 |
Paul Placeway | 8 | 115 | 44.97 |
William Russell | 9 | 4 | 4.40 |
Richard M. Schwartz | 10 | 2839 | 717.76 |
George Yu | 11 | 15 | 6.52 |