Scholarlink

We train Data2Vec2 models of various parameter scales on 300,000 h of unannotated Chinese multi-dialect and multi-accent speech data. These models are validated on multiple speech recognition datasets through fine-tuning and by utilizing them as feature extractors for CTC-based automatic speech recognition tasks. We are releasing these models to the open-source community to facilitate the research and application of speech processing technologies that support Chinese dialects and accents.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

TeleSpeechPT: Large-Scale Chinese Multi-dialect and Multi-accent Speech Pre-training

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hongjie Chen,
  • Zehan Li,
  • Guangmin Xia,
  • Boqing Liu,
  • Yan Yang,
  • Jian Kang,
  • Jie Li

摘要

We train Data2Vec2 models of various parameter scales on 300,000 h of unannotated Chinese multi-dialect and multi-accent speech data. These models are validated on multiple speech recognition datasets through fine-tuning and by utilizing them as feature extractors for CTC-based automatic speech recognition tasks. We are releasing these models to the open-source community to facilitate the research and application of speech processing technologies that support Chinese dialects and accents.