Minimum Bayes Risk Training of RNN-Transducer for End-to-End Speech Recognition
This work addresses speech recognition accuracy for Mandarin speech, offering incremental improvements over existing methods.
The paper tackles improving end-to-end speech recognition by proposing minimum Bayes risk (MBR) training for RNN-Transducer models, resulting in absolute character error rate reductions of 1.2% and 0.5% on read and spontaneous Mandarin speech over a strong baseline.
In this work, we propose minimum Bayes risk (MBR) training of RNN-Transducer (RNN-T) for end-to-end speech recognition. Specifically, initialized with a RNN-T trained model, MBR training is conducted via minimizing the expected edit distance between the reference label sequence and on-the-fly generated N-best hypothesis. We also introduce a heuristic to incorporate an external neural network language model (NNLM) in RNN-T beam search decoding and explore MBR training with the external NNLM. Experimental results demonstrate an MBR trained model outperforms a RNN-T trained model substantially and further improvements can be achieved if trained with an external NNLM. Our best MBR trained system achieves absolute character error rate (CER) reductions of 1.2% and 0.5% on read and spontaneous Mandarin speech respectively over a strong convolution and transformer based RNN-T baseline trained on ~21,000 hours of speech.