Unifying Streaming and Non-streaming Zipformer-based ASR

TL;DR

This paper introduces a unified end-to-end ASR model that effectively combines streaming and non-streaming capabilities using dynamic right-context and chunked attention, achieving high accuracy with controlled latency.

Contribution

It proposes a novel framework leveraging right-context in zipformer models to unify streaming and non-streaming ASR, reducing costs and improving flexibility.

Findings

Achieves 7.9% relative word error rate reduction.

Streaming performance approaches non-streaming accuracy with more right-context.

Flexible latency-accuracy tradeoff controlled by right-context size.

Abstract

There has been increasing interest in unifying streaming and non-streaming automatic speech recognition (ASR) models to reduce development, training, and deployment costs. We present a unified framework that trains a single end-to-end ASR model for both streaming and non-streaming applications, leveraging future context information. We propose to use dynamic right-context through the chunked attention masking in the training of zipformer-based ASR models. We demonstrate that using right-context is more effective in zipformer models compared to other conformer models due to its multi-scale nature. We analyze the effect of varying the number of right-context frames on accuracy and latency of the streaming ASR models. We use Librispeech and large in-house conversational datasets to train different versions of streaming and non-streaming models and evaluate them in a production grade server-client setup across diverse testsets of different domains. The proposed strategy reduces word error by relative 7.9\% with a small degradation in user-perceived latency. By adding more right-context frames, we are able to achieve streaming performance close to that of non-streaming models. Our approach also allows flexible control of the latency-accuracy tradeoff according to customers requirements.