| Paper ID | SPE-32.6 |
| Paper Title |
IMPROVING STREAMING AUTOMATIC SPEECH RECOGNITION WITH NON-STREAMING MODEL DISTILLATION ON UNSUPERVISED DATA |
| Authors |
Thibault Doutre, Wei Han, Min Ma, Zhiyun Lu, Chung-Cheng Chiu, Ruoming Pang, Arun Narayanan, Ananya Misra, Yu Zhang, Liangliang Cao, Google Inc., United States |
| Session | SPE-32: Speech Recognition 12: Self-supervised, Semi-supervised, Unsupervised Training |
| Location | Gather.Town |
| Session Time: | Thursday, 10 June, 13:00 - 13:45 |
| Presentation Time: | Thursday, 10 June, 13:00 - 13:45 |
| Presentation |
Poster
|
| Topic |
Speech Processing: [SPE-LVCR] Large Vocabulary Continuous Recognition/Search |
| IEEE Xplore Open Preview |
Click here to view in IEEE Xplore |
| Virtual Presentation |
Click here to watch in the Virtual Conference |
| Abstract |
Streaming end-to-end automatic speech recognition (ASR) models are widely used on smart speakers and on-device applications. Since these models are expected to transcribe speech with minimal latency, they are constrained to be causal with no future context, compared to their non-streaming counterparts. Consequently, streaming models usually perform worse than non-streaming models. We propose a novel and effective learning method by leveraging a non-streaming ASR model as a teacher to generate transcripts on an arbitrarily large data set, which is then used to distill knowledge into streaming ASR models. This way, we scale the training of streaming models to up to 3 million hours of YouTube audio. Experiments show that our approach can significantly reduce the word error rate (WER) of RNNT models not only on LibriSpeech but also on YouTube data in four languages. For example, in French, we are able to reduce the WER by 16.4% relatively to a baseline streaming model by leveraging a non-streaming teacher model trained on the same amount of labeled data as the baseline. |
