Semi-Autoregressive Streaming ASR With Label Context

arXiv:2309.10926v2 Announce Type: replace
Abstract: Non-autoregressive (NAR) modeling has gained significant interest in speech processing since these models achieve dramatically lower inference time than autoregressive (AR) models while also achieving good transcription accuracy. Since NAR automatic speech recognition (ASR) models must wait for the completion of the entire utterance before processing, some works explore streaming NAR models based on blockwise attention for low-latency applications. However, streaming NAR models significantly lag in accuracy compared to streaming AR and non-streaming NAR models. …

abstract accuracy arxiv asr automatic speech recognition context cs.cl cs.sd eess.as explore good inference modeling processing recognition speech speech processing speech recognition streaming transcription type