SC-MoE: Switch Conformer Mixture of Experts for Unified Streaming and Non-streaming Code-Switching ASR

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ye, Shuaishuai, Chen, Shunfei, Hu, Xinhui, Xu, Xinkang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916301591543808
author Ye, Shuaishuai
Chen, Shunfei
Hu, Xinhui
Xu, Xinkang
author_facet Ye, Shuaishuai
Chen, Shunfei
Hu, Xinhui
Xu, Xinkang
contents In this work, we propose a Switch-Conformer-based MoE system named SC-MoE for unified streaming and non-streaming code-switching (CS) automatic speech recognition (ASR), where we design a streaming MoE layer consisting of three language experts, which correspond to Mandarin, English, and blank, respectively, and equipped with a language identification (LID) network with a Connectionist Temporal Classification (CTC) loss as a router in the encoder of SC-MoE to achieve a real-time streaming CS ASR system. To further utilize the language information embedded in text, we also incorporate MoE layers into the decoder of SC-MoE. In addition, we introduce routers into every MoE layer of the encoder and the decoder and achieve better recognition performance. Experimental results show that the SC-MoE significantly improves CS ASR performances over baseline with comparable computational efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2406_18021
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle SC-MoE: Switch Conformer Mixture of Experts for Unified Streaming and Non-streaming Code-Switching ASR
Ye, Shuaishuai
Chen, Shunfei
Hu, Xinhui
Xu, Xinkang
Sound
Machine Learning
Audio and Speech Processing
In this work, we propose a Switch-Conformer-based MoE system named SC-MoE for unified streaming and non-streaming code-switching (CS) automatic speech recognition (ASR), where we design a streaming MoE layer consisting of three language experts, which correspond to Mandarin, English, and blank, respectively, and equipped with a language identification (LID) network with a Connectionist Temporal Classification (CTC) loss as a router in the encoder of SC-MoE to achieve a real-time streaming CS ASR system. To further utilize the language information embedded in text, we also incorporate MoE layers into the decoder of SC-MoE. In addition, we introduce routers into every MoE layer of the encoder and the decoder and achieve better recognition performance. Experimental results show that the SC-MoE significantly improves CS ASR performances over baseline with comparable computational efficiency.
title SC-MoE: Switch Conformer Mixture of Experts for Unified Streaming and Non-streaming Code-Switching ASR
topic Sound
Machine Learning
Audio and Speech Processing
url https://arxiv.org/abs/2406.18021