A novel pyramidal-FSMN architecture with lattice-free MMI for speech recognition
This work addresses speech recognition accuracy for applications like transcription, presenting incremental improvements over existing methods.
The paper tackled speech recognition by proposing a novel pyramidal-FSMN architecture with lattice-free MMI and joint training, achieving word error rates of 3.62% on Librispeech and 10.89% on Switchboard, with further improvement to 2.97% using RNNLM rescoring.
Deep Feedforward Sequential Memory Network (DFSMN) has shown superior performance on speech recognition tasks. Based on this work, we propose a novel network architecture which introduces pyramidal memory structure to represent various context information in different layers. Additionally, res-CNN layers are added in the front to extract more sophisticated features as well. Together with lattice-free maximum mutual information (LF-MMI) and cross entropy (CE) joint training criteria, experimental results show that this approach achieves word error rates (WERs) of 3.62% and 10.89% respectively on Librispeech and LDC97S62 (Switchboard 300 hours) corpora. Furthermore, Recurrent neural network language model (RNNLM) rescoring is applied and a WER of 2.97% is obtained on Librispeech.