Context dependant phone mapping for cross-lingual acoustic modeling
This paper presents a novel method for acoustic modeling with limited training data. The idea is to leverage on a well-trained acoustic model of a source language. In this paper, a conventional HMM/GMM triphone acoustic model of the source language is used to derive likelihood scores for each featur...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/97368 http://hdl.handle.net/10220/11891 |
| _version_ | 1811678537812279296 |
|---|---|
| author | Do, Van Hai Xiao, Xiong Chng, Eng Siong Li, Haizhou |
| author2 | School of Computer Engineering |
| author_facet | School of Computer Engineering Do, Van Hai Xiao, Xiong Chng, Eng Siong Li, Haizhou |
| author_sort | Do, Van Hai |
| collection | NTU |
| description | This paper presents a novel method for acoustic modeling with limited training data. The idea is to leverage on a well-trained acoustic model of a source language. In this paper, a conventional HMM/GMM triphone acoustic model of the source language is used to derive likelihood scores for each feature vector of the target language. These scores are then mapped to triphones of the target language using neural networks. We conduct a case study where Malay is the source language while English (Aurora-4 task) is the target language. Experimental results on the Aurora-4 (clean test set) show that by using only 7, 16, and 55 minutes of English training data, we achieve 21.58%, 17.97%, and 12.93% word error rate, respectively. These results outperform the conventional HMM/GMM and hybrid systems significantly. |
| first_indexed | 2024-10-01T02:54:51Z |
| format | Conference Paper |
| id | ntu-10356/97368 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T02:54:51Z |
| publishDate | 2013 |
| record_format | dspace |
| spelling | ntu-10356/973682020-05-28T07:17:25Z Context dependant phone mapping for cross-lingual acoustic modeling Do, Van Hai Xiao, Xiong Chng, Eng Siong Li, Haizhou School of Computer Engineering International Symposium on Chinese Spoken Language Processing (8th : 2012 : Kowloon, Hong Kong) Temasek Laboratories DRNTU::Engineering::Computer science and engineering This paper presents a novel method for acoustic modeling with limited training data. The idea is to leverage on a well-trained acoustic model of a source language. In this paper, a conventional HMM/GMM triphone acoustic model of the source language is used to derive likelihood scores for each feature vector of the target language. These scores are then mapped to triphones of the target language using neural networks. We conduct a case study where Malay is the source language while English (Aurora-4 task) is the target language. Experimental results on the Aurora-4 (clean test set) show that by using only 7, 16, and 55 minutes of English training data, we achieve 21.58%, 17.97%, and 12.93% word error rate, respectively. These results outperform the conventional HMM/GMM and hybrid systems significantly. 2013-07-18T07:23:25Z 2019-12-06T19:41:55Z 2013-07-18T07:23:25Z 2019-12-06T19:41:55Z 2012 2012 Conference Paper Do, V. H., Xiao, X., Chng, E. S., & Li, H. (2012). Context dependant phone mapping for cross-lingual acoustic modeling. 2012 8th International Symposium on Chinese Spoken Language Processing (ISCSLP). https://hdl.handle.net/10356/97368 http://hdl.handle.net/10220/11891 10.1109/ISCSLP.2012.6423496 en © 2012 IEEE. |
| spellingShingle | DRNTU::Engineering::Computer science and engineering Do, Van Hai Xiao, Xiong Chng, Eng Siong Li, Haizhou Context dependant phone mapping for cross-lingual acoustic modeling |
| title | Context dependant phone mapping for cross-lingual acoustic modeling |
| title_full | Context dependant phone mapping for cross-lingual acoustic modeling |
| title_fullStr | Context dependant phone mapping for cross-lingual acoustic modeling |
| title_full_unstemmed | Context dependant phone mapping for cross-lingual acoustic modeling |
| title_short | Context dependant phone mapping for cross-lingual acoustic modeling |
| title_sort | context dependant phone mapping for cross lingual acoustic modeling |
| topic | DRNTU::Engineering::Computer science and engineering |
| url | https://hdl.handle.net/10356/97368 http://hdl.handle.net/10220/11891 |
| work_keys_str_mv | AT dovanhai contextdependantphonemappingforcrosslingualacousticmodeling AT xiaoxiong contextdependantphonemappingforcrosslingualacousticmodeling AT chngengsiong contextdependantphonemappingforcrosslingualacousticmodeling AT lihaizhou contextdependantphonemappingforcrosslingualacousticmodeling |