Online Automatic Speech Recognition with Listen, Attend and Spell Model
This work addresses the problem of real-time speech recognition for Mandarin dictation, offering a production-scale deployment that is incremental but practical.
The paper tackled the limitations of attention-based ASR models in online operation by analyzing issues with silence regions and attention reliability, proposing a novel technique that achieves fully online recognition with a character error rate within 4% relative to offline models and 12% lower latency compared to hybrid models.
The Listen, Attend and Spell (LAS) model and other attention-based automatic speech recognition (ASR) models have known limitations when operated in a fully online mode. In this paper, we analyze the online operation of LAS models to demonstrate that these limitations stem from the handling of silence regions and the reliability of online attention mechanism at the edge of input buffers. We propose a novel and simple technique that can achieve fully online recognition while meeting accuracy and latency targets. For the Mandarin dictation task, our proposed approach can achieve a character error rate in online operation that is within 4% relative to an offline LAS model. The proposed online LAS model operates at 12% lower latency relative to a conventional neural network hidden Markov model hybrid of comparable accuracy. We have validated the proposed method through a production scale deployment, which, to the best of our knowledge, is the first such deployment of a fully online LAS model.