《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (4): 1104-1112.DOI: 10.11772/j.issn.1001-9081.2024030386

• 人工智能 • 上一篇    下一篇

基于多标签关系图和局部动态重构学习的多标签分类模型

胡婕1,2,3, 郑启扬1, 孙军1,2,3(), 张龑1,2,3   

  1. 1.湖北大学 计算机学院,武汉 430062
    2.大数据智能分析与行业应用湖北省重点实验室(湖北大学),武汉 430062
    3.智慧政务与人工智能应用湖北省工程研究中心(湖北大学),武汉 430062
  • 收稿日期:2024-04-08 修回日期:2024-07-04 接受日期:2024-07-17 发布日期:2024-10-12 出版日期:2025-04-10
  • 通讯作者: 孙军
  • 作者简介:胡婕(1977—),女,湖北汉川人,教授,博士,主要研究方向:复杂语义大数据管理、自然语言处理
    郑启扬(1997—),男,广东湛江人,硕士研究生,主要研究方向:自然语言处理
    孙军(1979—),女,湖北枣阳人,讲师,硕士,主要研究方向:自然语言处理
    张龑(1974—),男,湖北宜昌人,教授,博士,CCF会员,主要研究方向:软件工程、信息安全。
  • 基金资助:
    国家自然科学基金资助项目(61977021)

Multi-label classification model based on multi-label relational graph and local dynamic reconstruction learning

Jie HU1,2,3, Qiyang ZHENG1, Jun SUN1,2,3(), Yan ZHANG1,2,3   

  1. 1.School of Computer Science,Hubei University,Wuhan Hubei 430062,China
    2.Hubei Key Laboratory of Big Data Intelligent Analysis and Application (Hubei University),Wuhan Hubei 430062,China
    3.Engineering Research Center of Hubei Province in Intelligent Government Affairs and Application of Artificial Intelligence (Hubei University),Wuhan Hubei 430062,China
  • Received:2024-04-08 Revised:2024-07-04 Accepted:2024-07-17 Online:2024-10-12 Published:2025-04-10
  • Contact: Jun SUN
  • About author:HU Jie, born in 1977, Ph. D., professor. Her research interests include complex semantic big data management, natural language processing.
    ZHENG Qiyang, born in 1997, M. S. candidate. His research interests include natural language processing.
    SUN Jun, born in 1979, M. S., lecturer. Her research interests include natural language processing.
    ZHANG Yan, born in 1974, Ph. D., professor. His research interests include software engineering, information security.
  • Supported by:
    National Natural Science Foundation of China(61977021)

摘要:

在多标签分类任务中,现有模型对依赖关系的构建主要考虑标签在训练集中是否共现,而忽视了标签之间各种不同类型的关系以及在不同样本中的动态交互关系。因此,结合多标签关系图和局部动态重构图学习更完整的标签依赖关系。首先,根据标签的全局共现关系,采用数据驱动的方式构建多标签关系图,学习标签之间不同类型的依赖关系;其次,通过标签注意力机制探索文本信息和标签语义的关联性;最后,对标签图进行动态重构学习,以捕获标签之间的局部特定关系。在3个公开数据集BibTeX、Delicious和Reuters-21578上的实验结果表明,所提模型的宏平均F1(maF1)值相较于MrMP(Multi-relation Message Passing)分别提高了1.6、1.0和2.2个百分点,综合性能得到提升。

关键词: 多标签分类, 多标签关系图, 标签依赖关系, 局部动态重构图, 标签注意力机制

Abstract:

In multi-label classification tasks, the existing models mainly consider the co-occurrences of labels in the training set when constructing dependency, ignoring the various types of relationships and the dynamic interactions in different samples among labels. Therefore, the multi-label relational graph and the local dynamic reconstruction graph were combined to learn more complete label dependency. Firstly, based on the global co-occurrence relationships of labels, a multi-label relational graph was constructed using a data-driven approach to learn different types of dependency among labels. Secondly, the relevance of text information and label semantics was explored through the label attention mechanism. Finally, the label graph was reconstructed dynamically to learn and capture local-specific relationships among labels. Experiments were conducted on three public datasets BibTeX, Delicious, and Reuters-21578. The results show that the proposed model increases the macro average F1 (maF1) value by 1.6, 1.0 and 2.2 percentage points, respectively, and achieves improvements of comprehensive performance compared with Multi-relation Message Passing (MrMP).

Key words: multi-label classification, multi-label relational graph, label dependency, local dynamic reconstruction graph, label attention mechanism

中图分类号: