Language model fusion for streaming end to end speech recognition
This work addresses streaming and tail recognition problems for practical applications like dictation and voice search, representing an incremental improvement over existing fusion methods.
The paper tackled the challenges of streaming speech recognition and handling rare linguistic content by integrating a language model into a streaming Recurrent Neural Network Transducer, achieving up to an 8.5% reduction in word error rate.
Streaming processing of speech audio is required for many contemporary practical speech recognition tasks. Even with the large corpora of manually transcribed speech data available today, it is impossible for such corpora to cover adequately the long tail of linguistic content that's important for tasks such as open-ended dictation and voice search. We seek to address both the streaming and the tail recognition challenges by using a language model (LM) trained on unpaired text data to enhance the end-to-end (E2E) model. We extend shallow fusion and cold fusion approaches to streaming Recurrent Neural Network Transducer (RNNT), and also propose two new competitive fusion approaches that further enhance the RNNT architecture. Our results on multiple languages with varying training set sizes show that these fusion methods improve streaming RNNT performance through introducing extra linguistic features. Cold fusion works consistently better on streaming RNNT with up to a 8.5% WER improvement.