
软件学报 ISSN 1000-9825, CODEN RUXUEW E-mail: jos@iscas.ac.cn
Journal of Software,2020,31(4):10791089 [doi: 10.13328/j.cnki.jos.005923] http://www.jos.org.cn
©中国科学院软件研究所版权所有. Tel: +86-10-62562563
基于标签语义注意力的多标签文本分类
肖
琳
,
陈博理
,
黄
鑫
,
刘华锋
,
景丽萍
,
于
剑
(交通数据分析与挖掘北京市重点实验室(北京交通大学),北京 100044)
通信作者: 景丽萍, E-mail: lpjing@bjtu.edu.cn
摘 要: 自大数据蓬勃发展以来,多标签分类一直是令人关注的重要问题,在现实生活中有许多实际应用,如文本
分类、图像识别、视频注释、多媒体信息检索等.传统的多标签文本分类算法将标签视为没有语义信息的符号,然
而,在许多情况下,文本的标签是具有特定语义的,标签的语义信息和文档的内容信息是有对应关系的,为了建立两
者之间的联系并加以利用, 提出了一种基于标签语义注意力的多标签文本分类(LAbel Semantic Attention
Multi-label Classification,简称 LASA)方法,依赖于文档的文本和对应的标签,在文档和标签之间共享单词表示.对于
文档嵌入,使用双向长短时记忆(bi-directional long short-term memory,简称 Bi-LSTM)获取每个单词的隐表示,通过
使用标签语义注意力机制获得文档中每个单词的权重,从而考虑到每个单词对当前标签的重要性.另外,标签在语义
空间里往往是相互关联的,使用标签的语义信息同时也考虑了标签的相关性.在标准多标签文本分类的数据集上得
到的实验结果表明,所提出的方法能够有效地捕获重要的单词,并且其性能优于当前先进的多标签文本分类算法.
关键词: 多标签学习;文本分类;标签语义;注意力机制
中图法分类号: TP311
中文引用格式: 肖琳,陈博理,黄鑫,刘华锋,景丽萍,于剑.基于标签语义注意力的多标签文本分类.软件学报,2020,31(4):
10791089. http ://www.jos.org.cn/1000-9825/5923.htm
英文引用格式: Xiao L, Chen BL, Huang X, Liu HF, Jing LP, Yu J. Multi-label text classification method based on label semantic
information. Ruan Ji an Xue Bao/Jo urnal of Software, 2020,31 (4):10791089 (in Chin ese). http://www.jos.org.cn/1 000-9825/5923.
htm
Multi-label Text Classifica tion Metho d Based on Label Semanti c Informa tion
XIAO Lin, CHEN Bo-Li, HUANG Xin, LIU Hua-Feng, JING Li-Ping, YU Jian
(Beijing Key Laboratory of Traffic Data Analysis and Mining (Beijing Jiaotong University), Beijing 100044, China)
Abstra ct : Multi-label classification has been a practical and important problem since the boom of big data. There are many practical
applications, such as text classification, i mage recognition, vid eo annotation, multimedia information r etrieval, etc. Tr aditional multi-label
text classification algorithms regard l abels as symbols without inherent semantics. However, in many scenarios these labels have spe cific
semantics, and the semantic information of l abels have corresponding relati onship with the content information of the documents, in order
to establish the connection between them and make use of them, a label semantic attention multi-label classification (LASA) method is
proposed based on label semantic attention. The texts and labels of the document are relied on to share the word representation between
the texts and labels. For documents embedding, bi-directional long short-term memory (Bi-LSTM) is used to obtain the hidden
representation of each word. The weight of each word in the document is obtained by using the semantic representation of the label, thus
基金项目: 国家自然科学基金(61822601, 61773050, 61632004); 北京市自然科学基金(Z180006); 北京市科委项目(Z18110000
8918012)
Foundation item: National Natural Science Foundation of China (61822601, 61773050, 61632004); Beijing Natural Science
Foundation of China (Z180006); Beijing Municipal Science & Technology Commission (Z181100008918012)
本文由“非经典条件下的机器学习方法”专题特约编辑高新波教授、黎铭教授、李天瑞教授推荐.
收稿时间:
2019-05-29; 修改时间: 2019-07-29; 采用时间: 2019-09-20; jos 在线出版时间: 2020-01-10
CNKI 网络优先出版: 2020-01-14 09:53:23, http://kns.cnki.net/kcms/d etail/11.2560.TP.20200114.0953.009.html
评论