Transfer Learning of Language-independent End-to-end ASR with Language Model Fusion
Authors:
- 1 May 2019
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 6096-6100
Abstract
This work explores better adaptation methods to low-resource languages using an external language model (LM) under the framework of transfer learning. We first build a language-independent ASR system in a unified sequence-to-sequence (S2S) architecture with a shared vocabulary among all languages. During adaptation, we perform LM fusion transfer, where an external LM is integrated into the decoder network of the attention-based S2S model in the whole adaptation stage, to effectively incorporate linguistic context of the target language. We also investigate various seed models for transfer learning. Experimental evaluations using the IARPA BABEL data set show that LM fusion transfer improves performances on all target five languages compared with simple transfer learning when the external text data is available. Our final system drastically reduces the performance gap from the hybrid systems.Keywords
All Related Versions
This publication has 19 references indexed in Scilit:
- Cross-lingual adaptation of a CTC-based multilingual acoustic modelSpeech Communication, 2018
- Sequence-Based Multi-Lingual Low Resource Speech Recognition2018
- Language independent end-to-end architecture for joint language identification and speech recognition2017
- Hybrid CTC/Attention Architecture for End-to-End Speech RecognitionIEEE Journal of Selected Topics in Signal Processing, 2017
- Using Target-side Monolingual Data for Neural Machine Translation through Multi-task Learning2017
- Listen, attend and spell: A neural network for large vocabulary conversational speech recognition2016
- Multilingual deep neural network based acoustic modeling for rapid language adaptation2014
- The language-independent bottleneck features2012
- Connectionist temporal classification2006
- Long Short-Term MemoryNeural Computation, 1997