Exploiting Low-Resource Code-Switching Data to Mandarin-English Speech Recognition Systems

1070

Authors

nan

Abstract

AbstractIn this paper, we investigate how to use limited code-switching data to implement a code-switching speech recognition system. We utilize the Transformer end-to-end model to develop our code switching speech recognition system, which is trained with the Mandarin dataset and a small amount of Mandarin-English code switching dataset, as the baseline of this paper. Next, we compare the performance of systems after adding multi-task learning and transfer learning. Character Error Rate(CER) is adopted as the criterion for the system. Finally, we combined the three systems with the language model, respectively, our best result dropped to 23.9% compared with the baseline of 28.7%.

Keywords

nan

Full Content

�� The 33rd Conference on Computational Linguistics and Speech Processing (ROCLING 2021) Taoyuan, Taiwan, October 15-16, 2021. The Association for Computational Linguistics and Chinese Language Processing 81 �� Xe � � � � � � �� Y [1 : u] = Y [1], ..., Y [u]�� p��(Y |Xe) �� [p��(Y [2]|Y [1], Xe), ..., p��(Y [u + 1]|Y [1 : u], Xe)] = ��(ZdW�� + b��) p��(Y |Xe) = � u p��(Y [u + 1]|Y [1 : u], Xe) � � Zd � � � � � � ��W�� ∈ Rd��×d��, b�� ∈ Rd�� d�� …

Content truncated to 500 words

Labeled Text

Publication Details

Paper ID:: 1070
Year:: 2021
Publication:: TIPSTER
Publication Type:: conf
DOI:: nan

Links

Paper URL PDF Publication URL