FastEmit: Low-latency Streaming ASR with Sequence-level Emission Regularization
This work addresses the challenge of low-latency speech recognition for real-time applications, offering a novel method that improves upon existing techniques.
The paper tackles the problem of reducing latency in streaming automatic speech recognition without degrading accuracy, achieving a latency reduction of 150-300 ms with better accuracy on a Voice Search test set and improving WER from 4.4%/8.9% to 3.1%/7.5% while reducing 90th percentile latency from 210 ms to 30 ms on LibriSpeech.
Streaming automatic speech recognition (ASR) aims to emit each hypothesized word as quickly and accurately as possible. However, emitting fast without degrading quality, as measured by word error rate (WER), is highly challenging. Existing approaches including Early and Late Penalties and Constrained Alignments penalize emission delay by manipulating per-token or per-frame probability prediction in sequence transducer models. While being successful in reducing delay, these approaches suffer from significant accuracy regression and also require additional word alignment information from an existing model. In this work, we propose a sequence-level emission regularization method, named FastEmit, that applies latency regularization directly on per-sequence probability in training transducer models, and does not require any alignment. We demonstrate that FastEmit is more suitable to the sequence-level optimization of transducer models for streaming ASR by applying it on various end-to-end streaming ASR networks including RNN-Transducer, Transformer-Transducer, ConvNet-Transducer and Conformer-Transducer. We achieve 150-300 ms latency reduction with significantly better accuracy over previous techniques on a Voice Search test set. FastEmit also improves streaming ASR accuracy from 4.4%/8.9% to 3.1%/7.5% WER, meanwhile reduces 90th percentile latency from 210 ms to only 30 ms on LibriSpeech.