HMM-Free Encoder Pre-Training for Streaming RNN Transducer
This work addresses the challenge of training efficient streaming speech recognition models without relying on HMM-based alignments, making it lexicon-free and applicable to new languages.
The paper tackles the performance gap between streaming and non-streaming RNN-T models by proposing an HMM-free encoder pre-training method using frame-wise labels generated from CTC model spikes, which reduces WER by 5-11% and emission latency by 60 ms on LibriSpeech and MLS English tasks.
This work describes an encoder pre-training procedure using frame-wise label to improve the training of streaming recurrent neural network transducer (RNN-T) model. Streaming RNN-T trained from scratch usually performs worse than non-streaming RNN-T. Although it is common to address this issue through pre-training components of RNN-T with other criteria or frame-wise alignment guidance, the alignment is not easily available in end-to-end manner. In this work, frame-wise alignment, used to pre-train streaming RNN-T's encoder, is generated without using a HMM-based system. Therefore an all-neural framework equipping HMM-free encoder pre-training is constructed. This is achieved by expanding the spikes of CTC model to their left/right blank frames, and two expanding strategies are proposed. To our best knowledge, this is the first work to simulate HMM-based frame-wise label using CTC model for pre-training. Experiments conducted on LibriSpeech and MLS English tasks show the proposed pre-training procedure, compared with random initialization, reduces the WER by relatively 5%~11% and the emission latency by 60 ms. Besides, the method is lexicon-free, so it is friendly to new languages without manually designed lexicon.