Abstract | ||
---|---|---|
This paper describes the BYBLOS system that BBN used to participate in the 2001 NIST Hub-5 evaluation benchmark. We outline the procedure used for training and decoding, and present the algorithmic improvements made to the system, along with experimental results. These improvements include a Gaussian splitting initialization procedure, the use of Linear Discriminant Analysis, and processing of additional acoustic training data. We also discuss our system combination and confidence-based thresholding methods. Experiments on an internal validation test set show that all these system improvements provide a 8.1% relative reduction in word error rate compared to our 2000 LVCSR system. |
Year | DOI | Venue |
---|---|---|
2002 | 10.1109/ICASSP.2002.5743819 | ICASSP |
Field | DocType | Volume |
Computer science,Word error rate,Speech recognition,NIST,Natural language processing,Artificial intelligence,Initialization,Decoding methods,Linear discriminant analysis,Thresholding,Vocabulary,Test set | Conference | 1 |
ISSN | Citations | PageRank |
1520-6149 | 3 | 0.38 |
References | Authors | |
1 | 8 |
Name | Order | Citations | PageRank |
---|---|---|---|
Spyros Matsoukas | 1 | 449 | 37.56 |
Thomas Colthurst | 2 | 78 | 7.71 |
Owen Kimball | 3 | 83 | 17.82 |
Alex Solomonoff | 4 | 425 | 37.68 |
Fred Richardson | 5 | 179 | 18.80 |
Carl Quillen | 6 | 41 | 4.94 |
Herbert Gish | 7 | 447 | 100.85 |
Pierre L. Dognin | 8 | 31 | 10.24 |