Improving Code-Switching Speech Recognition with TTS Data Augmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yeo, Yue Heng, Hu, Yuchen, Gopal, Shreyas, Peng, Yizhou, Liu, Hexin, Chng, Eng Siong
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915704798707712
author Yeo, Yue Heng
Hu, Yuchen
Gopal, Shreyas
Peng, Yizhou
Liu, Hexin
Chng, Eng Siong
author_facet Yeo, Yue Heng
Hu, Yuchen
Gopal, Shreyas
Peng, Yizhou
Liu, Hexin
Chng, Eng Siong
contents Automatic speech recognition (ASR) for conversational code-switching speech remains challenging due to the scarcity of realistic, high-quality labeled speech data. This paper explores multilingual text-to-speech (TTS) models as an effective data augmentation technique to address this shortage. Specifically, we fine-tune the multilingual CosyVoice2 TTS model on the SEAME dataset to generate synthetic conversational Chinese-English code-switching speech, significantly increasing the quantity and speaker diversity of available training data. Our experiments demonstrate that augmenting real speech with synthetic speech reduces the mixed error rate (MER) from 12.1 percent to 10.1 percent on DevMan and from 17.8 percent to 16.0 percent on DevSGE, indicating consistent performance gains. These results confirm that multilingual TTS is an effective and practical tool for enhancing ASR robustness in low-resource conversational code-switching scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2601_00935
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Improving Code-Switching Speech Recognition with TTS Data Augmentation
Yeo, Yue Heng
Hu, Yuchen
Gopal, Shreyas
Peng, Yizhou
Liu, Hexin
Chng, Eng Siong
Audio and Speech Processing
Artificial Intelligence
Automatic speech recognition (ASR) for conversational code-switching speech remains challenging due to the scarcity of realistic, high-quality labeled speech data. This paper explores multilingual text-to-speech (TTS) models as an effective data augmentation technique to address this shortage. Specifically, we fine-tune the multilingual CosyVoice2 TTS model on the SEAME dataset to generate synthetic conversational Chinese-English code-switching speech, significantly increasing the quantity and speaker diversity of available training data. Our experiments demonstrate that augmenting real speech with synthetic speech reduces the mixed error rate (MER) from 12.1 percent to 10.1 percent on DevMan and from 17.8 percent to 16.0 percent on DevSGE, indicating consistent performance gains. These results confirm that multilingual TTS is an effective and practical tool for enhancing ASR robustness in low-resource conversational code-switching scenarios.
title Improving Code-Switching Speech Recognition with TTS Data Augmentation
topic Audio and Speech Processing
Artificial Intelligence
url https://arxiv.org/abs/2601.00935