Multilingual sequence-to-sequence speech recognition: architecture, transfer learning, and language modeling
This work addresses low-resource speech recognition for multilingual applications, representing an incremental improvement in transfer learning methods.
The paper tackled low-resource automatic speech recognition by building a multilingual sequence-to-sequence model using data from 10 languages and transferring it to 4 other languages, achieving substantial gains over monolingual models and improving word error rate with a recurrent neural network language model to match performance with double the training data.
Sequence-to-sequence (seq2seq) approach for low-resource ASR is a relatively new direction in speech research. The approach benefits by performing model training without using lexicon and alignments. However, this poses a new problem of requiring more data compared to conventional DNN-HMM systems. In this work, we attempt to use data from 10 BABEL languages to build a multi-lingual seq2seq model as a prior model, and then port them towards 4 other BABEL languages using transfer learning approach. We also explore different architectures for improving the prior multilingual seq2seq model. The paper also discusses the effect of integrating a recurrent neural network language model (RNNLM) with a seq2seq model during decoding. Experimental results show that the transfer learning approach from the multilingual model shows substantial gains over monolingual models across all 4 BABEL languages. Incorporating an RNNLM also brings significant improvements in terms of %WER, and achieves recognition performance comparable to the models trained with twice more training data.