Chinese short news text classification based on BERT and sparse autoencoder

Jiuzhou Lin

doi:10.1117/12.2659618

30 November 2022 Chinese short news text classification based on BERT and sparse autoencoder

Jiuzhou Lin

Proceedings Volume 12456, International Conference on Artificial Intelligence and Intelligent Information Processing (AIIIP 2022); 124562N (2022) https://doi.org/10.1117/12.2659618
Event: International Conference on Artificial Intelligence and Intelligent Information Processing (AIIIP 2022), 2022, Qingdao, China

Abstract

Short news text classification plays an import role in natural language processing as the popularity of mobile phones. In this paper we propose a Chinese short news text classification method based on BERT and sparse autoencoder, regarding the overfitting caused by pretrained BERT. We use the BERT for text representation, the output vectors of BERT are dimension reduced through the sparse autoencoder, and then the Softmax classifier takes the reduced vectors as input to get the prediction of the input text. Experimental results show that our method mitigate the unbalance of the performance of different categories, raises the overall classification performance by six percentage, effectively alleviates the overfitting of text representation of BERT, and achieve a better Chinese short text classification performance than using naïve autoencoder and without autoencoder.

Citation Download Citation

Jiuzhou Lin "Chinese short news text classification based on BERT and sparse autoencoder", Proc. SPIE 12456, International Conference on Artificial Intelligence and Intelligent Information Processing (AIIIP 2022), 124562N (30 November 2022); https://doi.org/10.1117/12.2659618

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available