Abstract | ||
---|---|---|
In this paper we propose a novel way of estimating confidences for words that are recognized by a speech recognition system, together with a natural methodology for evaluating the overall quality of those confidence estimates. Our approach is based on an interpretation of a confidence as the probability that the corresponding recognized word is correct, and makes use of generalized linear models as a means for combining various predictor scores so as to arrive at confidence estimates. Experimental results using these models are presented based on four different sources of speech data: switchboard, Spanish and Mandarin CallHome, and Wall Street Journal |
Year | DOI | Venue |
---|---|---|
1997 | 10.1109/ICASSP.1997.596076 | Acoustics, Speech, and Signal Processing, 1997. ICASSP-97., 1997 IEEE International Conference |
Keywords | Field | DocType |
estimation theory,prediction theory,probability,speech recognition,Mandarin CallHome speech,Wall Street Journal speech,confidence estimation,confidence evaluation,generalized linear models,predictor scores,probabilistic approach,quality,recognized word,speech data,speech recognition system,switchboard speech | Pattern recognition,Computer science,Speech recognition,Generalized linear model,Natural language processing,Artificial intelligence,Probabilistic logic,Estimation theory,Mandarin Chinese | Conference |
Volume | ISSN | ISBN |
2 | 1520-6149 | 0-8186-7919-0 |
Citations | PageRank | References |
49 | 4.30 | 7 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Gillick, L. | 1 | 49 | 4.30 |
Ito, Y. | 2 | 54 | 6.22 |
Young, J. | 3 | 186 | 13.76 |