Name
Abstract | ||
|---|---|---|
Lack of data is a problem in training language models for conversational speech recognition, particularly for languages other than English. Experiments in English have successfully used web-based text collection targeted for a conversational style to augment small sets of transcribed speech; here we look at extending these techniques to Mandarin. In addition, we investigate different techniques for topic adaptation. Experiments in recognizing Mandarin telephone conversations show that use of filtered web data leads to a 28% reduction in perplexity and 7% reduction in character error rate, with most of the gain due to the general filtered web data. |
Year | DOI | Venue |
|---|---|---|
2005 | 10.1109/ICASSP.2005.1415182 | ICASSP '05). IEEE International Conference |
Keywords | Field | DocType |
Internet,error statistics,learning (artificial intelligence),natural languages,speech recognition,text analysis,Mandarin conversational speech recognition,Mandarin speech recognition,Mandarin telephone conversations,Web-based text collection,Web-data augmented language models,character error rate,language model training,perplexity,topic adaptation | Perplexity,Computer science,Word error rate,Speech recognition,Natural language,Artificial intelligence,Natural language processing,Information engineering,Telephony,Mandarin Chinese,Language model,The Internet | Conference |
Volume | ISSN | ISBN |
1 | 1520-6149 | 0-7803-8874-7 |
Citations | PageRank | References |
25 | 1.75 | 7 |
Authors | ||
6 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Tim Ng | 1 | 122 | 9.38 |
| Mari Ostendorf | 2 | 2462 | 348.75 |
| Mei-Yuh Hwang | 3 | 477 | 124.33 |
| Manhung Siu | 4 | 464 | 61.40 |
| Ivan Bulyko | 5 | 249 | 22.40 |
| Xin Lei | 6 | 219 | 19.36 |