基于拼音约束联合学习的汉语语音识别(Chinese Speech Recognition Based on Pinyin Constraint Joint Learning)

Renfeng Liang (梁仁凤), Zhengtao Yu (余正涛), Shengxiang Gao (高盛祥), Yuxin Huang (黄于欣), Junjun Guo (郭军军), Shuli Xu (许树理)


Abstract
当前的语音识别模型在英语、法语等表音文字中已经取得很好的效果。然而,汉语是 一种典型的表意文字,汉字与语音没有直接的对应关系,但拼音作为汉字读音的标注 符号,与汉字存在相互转换的内在联系。因此,在汉语语音识别中利用拼音作为解码 约束,引入一种更接近语音的归纳偏置。基于多任务学习框架,提出一种基于拼音约 束联合学习的汉语语音识别方法,以端到端的汉字语音识别为主任务,以拼音语音识 别为辅助任务,通过共享编码器,同时利用汉字与拼音识别结果作为监督信号,增强 编码器对汉语语音的表达能力。实验结果表明,相比基线模型,提出方法取得更优的 识别效果,词错误率WER降低了2.24个百分点
Anthology ID:
2020.ccl-1.70
Volume:
Proceedings of the 19th Chinese National Conference on Computational Linguistics
Month:
October
Year:
2020
Address:
Haikou, China
Editors:
Maosong Sun (孙茂松), Sujian Li (李素建), Yue Zhang (张岳), Yang Liu (刘洋)
Venue:
CCL
SIG:
Publisher:
Chinese Information Processing Society of China
Note:
Pages:
754–760
Language:
Chinese
URL:
https://aclanthology.org/2020.ccl-1.70
DOI:
Bibkey:
Cite (ACL):
Renfeng Liang, Zhengtao Yu, Shengxiang Gao, Yuxin Huang, Junjun Guo, and Shuli Xu. 2020. 基于拼音约束联合学习的汉语语音识别(Chinese Speech Recognition Based on Pinyin Constraint Joint Learning). In Proceedings of the 19th Chinese National Conference on Computational Linguistics, pages 754–760, Haikou, China. Chinese Information Processing Society of China.
Cite (Informal):
基于拼音约束联合学习的汉语语音识别(Chinese Speech Recognition Based on Pinyin Constraint Joint Learning) (Liang et al., CCL 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.ccl-1.70.pdf