Unifying Streaming and Non-streaming Zipformer-based ASR

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sharma, Bidisha, Durai, Karthik Pandia, Venkatesan, Shankar, Prakash, Jeena J, Kumar, Shashi, Chetlur, Malolan, Stolcke, Andreas
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911009591001088
author Sharma, Bidisha
Durai, Karthik Pandia
Venkatesan, Shankar
Prakash, Jeena J
Kumar, Shashi
Chetlur, Malolan
Stolcke, Andreas
author_facet Sharma, Bidisha
Durai, Karthik Pandia
Venkatesan, Shankar
Prakash, Jeena J
Kumar, Shashi
Chetlur, Malolan
Stolcke, Andreas
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2506_14434
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Unifying Streaming and Non-streaming Zipformer-based ASR
Sharma, Bidisha
Durai, Karthik Pandia
Venkatesan, Shankar
Prakash, Jeena J
Kumar, Shashi
Chetlur, Malolan
Stolcke, Andreas
Sound
Artificial Intelligence
Audio and Speech Processing
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.
title Unifying Streaming and Non-streaming Zipformer-based ASR
topic Sound
Artificial Intelligence
Audio and Speech Processing
url https://arxiv.org/abs/2506.14434