Recent advances in the IBM GALE Mandarin transcription system Export

@inproceedings{citeulike:2986224, abstract = {This paper describes the system and algorithmic developments in the automatic transcription of Mandarin broadcast speech made at IBM in the second year of the DARPA GALE program. Technical advances over our previous system include improved acoustic models using embedded tone modeling, and a new topic-adaptive language model (LM) rescoring technique based on dynamically generated LMs. We present results on three community-defined test sets designed to cover both the broadcast news and the broadcast conversation domain. It is shown that our new baseline system attains a 15.4\% relative reduction in character error rate compared with our previous GALE evaluation system. And a further 13.6\% improvement over the baseline is achieved with the two described techniques.}, author = {Chu, S. M. and Kuo, Hong-Kwang and Mangu, L. and Liu, Yi and Qin, Yong and Shi, Qin and Zhang, Shi L. and Aronowitz, H.}, booktitle = {Acoustics, Speech and Signal Processing, 2008. ICASSP 2008. IEEE International Conference on}, citeulike-article-id = {2986224}, citeulike-linkout-0 = {http://dx.doi.org/10.1109/ICASSP.2008.4518613}, citeulike-linkout-1 = {http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4518613}, doi = {10.1109/ICASSP.2008.4518613}, journal = {Acoustics, Speech and Signal Processing, 2008. ICASSP 2008. IEEE International Conference on}, keywords = {asr, gale, ibm}, pages = {4329--4332}, posted-at = {2008-07-11 01:21:38}, priority = {0}, title = {Recent advances in the IBM GALE Mandarin transcription system}, url = {http://dx.doi.org/10.1109/ICASSP.2008.4518613}, year = {2008} }

TY - CONF ID - citeulike:2986224 L3 - citeulike-article-id:2986224 N2 - This paper describes the system and algorithmic developments in the automatic transcription of Mandarin broadcast speech made at IBM in the second year of the DARPA GALE program. Technical advances over our previous system include improved acoustic models using embedded tone modeling, and a new topic-adaptive language model (LM) rescoring technique based on dynamically generated LMs. We present results on three community-defined test sets designed to cover both the broadcast news and the broadcast conversation domain. It is shown that our new baseline system attains a 15.4% relative reduction in character error rate compared with our previous GALE evaluation system. And a further 13.6% improvement over the baseline is achieved with the two described techniques. JF - Acoustics, Speech and Signal Processing, 2008. ICASSP 2008. IEEE International Conference on EP - 4332 TI - Recent advances in the IBM GALE Mandarin transcription system T2 - Acoustics, Speech and Signal Processing, 2008. ICASSP 2008. IEEE International Conference on SP - 4329 KW - asr KW - gale KW - ibm AU - Chu, SM AU - Kuo, Hong-Kwang AU - Mangu, L AU - Liu, Yi AU - Qin, Yong AU - Shi, Qin AU - Zhang, Shi AU - Aronowitz, H PY - 2008/// UR - http://dx.doi.org/10.1109/ICASSP.2008.4518613 ER -