AUTOMATIC PRONUNCIATION VERIFICATION FOR SPEECH RECOGNITION

Pronunciations for words are a critical component in an automated speech recognition system (ASR) as mis-recognitions may be caused by missing or inaccurate pronunciations. The need for high quality pronunciations has recently motivated data-driven techniques to gen-erate them [1]. We propose a data-driven and language-independent framework for verification of such pronunciations to further improve the lexicon quality in ASR. New candidate pronunciations are veri-fied by re-recognizing historical audio logs and examining the asso-ciated recognition costs. We build an additional pronunciation qual-ity feature from word and pronunciation frequencies in logs. A ma-chine learned classifier trained on these features achieves nearly 90% accuracy in labeling good vs bad pronunciations across all languages we tested. New pronunciations verified as good may be added to a dictionary, while bad pronunciations may be discarded or sent to experts for further evaluation. We simultaneously verify 5,000 to 30,000 new pronunciations within a few hours and show improve-ments in the ASR performance as a result of including pronuncia-tions verified by this system. 1