A new approach to text feature extraction using BiReGU and capsule networks

JiaQi Hao

doi:10.1117/12.2646320

23 August 2022 A new approach to text feature extraction using BiReGU and capsule networks

JiaQi Hao

Proceedings Volume 12330, International Conference on Cyber Security, Artificial Intelligence, and Digital Economy (CSAIDE 2022); 1233005 (2022) https://doi.org/10.1117/12.2646320
Event: International Conference on Cyber Security, Artificial Intelligence, and Digital Economy (CSAIDE 2022), 2022, Huzhou, China

Abstract

Traditional single-model text feature representation methods lack multi-level representation of features and cannot extract global and local information well. To this end, the text uses a two-layer ReGU network (BiReGU) and a capsule network to extract the global semantic and local semantic features of the text respectively, combining attention mechanisms and interaction forms to obtain a more comprehensive multi-level feature representation of the text. In addition, the traditional capsule network dynamic routing algorithm suffers from the problem of noisy capsule interference. This paper proposes to reduce the weight of noisy capsules based on the attention mechanism and Capsule's feature extraction method, thus reducing the interference to subsequent capsules. Comparative experiments are conducted on a generic dataset and the experimental results demonstrate the effectiveness of the proposed method.

Citation Download Citation

JiaQi Hao "A new approach to text feature extraction using BiReGU and capsule networks", Proc. SPIE 12330, International Conference on Cyber Security, Artificial Intelligence, and Digital Economy (CSAIDE 2022), 1233005 (23 August 2022); https://doi.org/10.1117/12.2646320

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