Task-specific adaptation of speech recognition models

A. Sankar1, A. Kannan1, B. Shahshahani1, E. Jackson2,3
1Nuance Communications, Menlo Park, CA, USA
2Nuance Communications, Menlo Park, CA
3Google, Mountain View, CA

Tóm tắt

Most published adaptation research focuses on speaker adaptation, and on adaptation for noisy channels and background environments. We study acoustic, grammar, and combined acoustic and grammar adaptation for creating task-specific recognition models. Comprehensive experimental results are presented using data from natural language quotes and a trading application. The results show that task adaptation gives substantial improvements in both utterance understanding accuracy, and recognition speed.

Từ khóa

#Speech recognition #Hidden Markov models #Distributed computing #Loudspeakers #Acoustic applications #Adaptation model #Smoothing methods #Acoustic noise #Background noise #Working environment noise

Tài liệu tham khảo

10.1109/ICASSP.1997.596223 10.1109/89.279278 chang, 1999, Improving Rejection with Semantic Slot-Based Confidence Scores, Proc EUROSPEECH, 271 10.1006/dspr.1999.0361 10.1109/89.506931 10.1109/89.496215 leggetter, 1995, Flexible speaker adaptation using maximum likelihood linear regression, Proc Spoken Lang Systems Technology Work shop, 110