System and method for discriminative pronunciation modeling for voice search

摘要

Disclosed herein are systems, computer-implemented methods, and computer-readable media for speech recognition. The method includes receiving speech utterances, assigning a pronunciation weight to each unit of speech in the speech utterances, each respective pronunciation weight being normalized at a unit of speech level to sum to 1, for each received speech utterance, optimizing the pronunciation weight by (1) identifying word and phone alignments and corresponding likelihood scores, and (2) discriminatively adapting the pronunciation weight to minimize classification errors, and recognizing additional received speech utterances using the optimized pronunciation weights. A unit of speech can be a sentence, a word, a context-dependent phone, a context-independent phone, or a syllable. The method can further include discriminatively adapting pronunciation weights based on an objective function. The objective function can be maximum mutual information (MMI), maximum likelihood (MLE) training, minimum classification error (MCE) training, or other functions known to those of skill in the art. Speech utterances can be names. The speech utterances can be received as part of a multimodal search or input. The step of discriminatively adapting pronunciation weights can further include stochastically modeling pronunciations.