Contextualized Automatic Speech Recognition with Dynamic Vocabulary Prediction and Activation
This addresses accuracy issues in ASR for applications requiring precise contextual phrase recognition, representing an incremental improvement over existing deep biasing methods.
The paper tackles the problem of maintaining contextual phrase integrity in automatic speech recognition by proposing an encoder-based phrase-level method with dynamic vocabulary prediction and activation, achieving relative WER reductions of 28.31% and 23.49% on datasets and reducing WER on contextual phrases by 72.04% and 75.69%.
Deep biasing improves automatic speech recognition (ASR) performance by incorporating contextual phrases. However, most existing methods enhance subwords in a contextual phrase as independent units, potentially compromising contextual phrase integrity, leading to accuracy reduction. In this paper, we propose an encoder-based phrase-level contextualized ASR method that leverages dynamic vocabulary prediction and activation. We introduce architectural optimizations and integrate a bias loss to extend phrase-level predictions based on frame-level outputs. We also introduce a confidence-activated decoding method that ensures the complete output of contextual phrases while suppressing incorrect bias. Experiments on Librispeech and Wenetspeech datasets demonstrate that our approach achieves relative WER reductions of 28.31% and 23.49% compared to baseline, with the WER on contextual phrases decreasing relatively by 72.04% and 75.69%.