LAR-SiCo: recommending law articles based on multi-label text classification

How to automatically recommend the law articles violated by the illegal advertisements is one of the hot topics in the smart justice. Because an illegal advertisement may violate one or more law articles, we proposed to consider the automatic recommendation of law articles as a multi-label text clas...

Full description

Saved in:
Bibliographic Details
Published in:International journal of machine learning and cybernetics Vol. 16; no. 5; pp. 3927 - 3941
Main Authors: Zhao, Hua, Li, Xiaoqian, Zeng, Qingtian, Zou, Zhenqi, Liang, Jinguo
Format: Journal Article
Language:English
Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.06.2025
Springer Nature B.V
Subjects:
Advertising
Artificial Intelligence
Classification
Complex Systems
Computational Intelligence
Control
Datasets
Engineering
Labels
Mechatronics
Natural language processing
Neural networks
Original Article
Pattern Recognition
Robotics
Semantics
Systems Biology
Text categorization
Smart justice
Multi-label text classification
Law articles recommendation
Illegal advertisement
ISSN:1868-8071, 1868-808X
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:How to automatically recommend the law articles violated by the illegal advertisements is one of the hot topics in the smart justice. Because an illegal advertisement may violate one or more law articles, we proposed to consider the automatic recommendation of law articles as a multi-label text classification (MLTC) task, and created a law article recommendation method based on multi-label text classification. To address the issue of insufficient labeled data in the field of illegal advertisement, a law article recommendation method based on label semantic information and label co-occurrence (LAR-SiCo) was proposed. LAR-SiCo firstly used the BERT to obtain the initial law article vector; secondly, based on Node2Vec and Graph convolution Network (GCN), the initial law article vector was updated iteratively by utilizing law article semantic information and law article co-occurrence, which was optimized using the proposed label loss. Finally, the classification loss and label loss were combined to optimize LAR- SiCo. At the same time, we constructed an illegal advertisement dataset for law articles recommendation. The experimental results showed that LAR-SiCo was feasible and effective in recommending the law articles violated by the illegal advertisements.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1868-8071
1868-808X
DOI:10.1007/s13042-024-02489-6