语音识别模型的训练方法及装置、电子设备和存储介质

By pre-training and splicing compressed nonlinear modules and incremental training, the recognition accuracy of the speech recognition model in multilingual and technical terminology scenarios was improved, solving the problem of insufficient model adaptability and achieving efficient domain-adaptive training.

CN116994586BActive Publication Date: 2026-07-17BEIJING YUNSIZHIXUE TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
BEIJING YUNSIZHIXUE TECH CO LTD
Filing Date
2023-08-08
Publication Date
2026-07-17

Smart Images

  • Figure CN116994586B_ABST
    Figure CN116994586B_ABST
Patent Text Reader

Abstract

本公开提供了一种语音识别模型的训练方法及装置、电子设备和存储介质，利用预训练语料对语音识别模型进行训练，得到已训练语音识别模型；将压缩非线性模块与所述已训练语音识别模型进行拼接；基于拼接后的已训练语音识别模型，对所述压缩非线性模块进行训练，不同所述压缩非线性模块用于拟合不同领域的语料。与相关技术相比，本公开实施例通过对不同领域对应的压缩非线性模块的训练，提高了不同的压缩非线性模块对语料的拟合能力；进而能够提升语音识别模型对于不同领域语料的识别能力及准确性。
Need to check novelty before this filing date? Find Prior Art