融合领域知识图谱的跨境民族文本分类方法  

Cross-border Ethnic Text Classification Method Integrating Domain Knowledge Map

作　　者：陈春吉 毛存礼[1,2] 雷雄丽[3] 满志博 陆杉 张勇丙 CHEN Chunji;MAO Cunli;LEI Xiongli;MAN Zhibo;LU Shan;ZHANG Yongbing(School of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China;Yunnan Key Laboratory of Artificial Intelligence,Kunming University of Science and Technology,Kunming 650500,China;School of Architecture and Art,Kunming Metallurgical College,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500 [2]昆明理工大学云南省人工智能重点实验室,云南昆明650500 [3]昆明冶金高等专科学校建筑与艺术学院,云南昆明650000

出　　处：《山西大学学报（自然科学版）》2022年第4期884-893,共10页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(61732005,61866019,61761026,61972186);云南省应用基础研究计划重点项目(2019FA023);云南省中青年学术和技术带头人后备人才项目(2019HB006);云南特色产业数字化研究与应用示范(202002AD080001)。

摘　　要：跨境民族文本分类任务是跨境民族文化分析中的基础性工作,其目的是将跨境民族文化文本进行归类处理。针对跨境民族文化数据分类面临类别交叉的问题,提出融合领域知识图谱的跨境民族文本分类方法,利用跨境民族文化知识图谱对文本中的跨境民族实体进行语义扩展,通过实体在知识图谱中的类别特征来增强文本的类别语义特征。此外,通过掩码自注意力机制分别对文本的词级、句子级进行特征提取以此得到文本中句子的局部特征和全局特征。实验表明,本文方法在跨境民族文化数据集中相比基线模型的F1值提升了11.9%。The task of cross-border ethnic text classification is the basic work in cross-border ethnic cultural analysis,and its purpose is to classify cross-border ethnic cultural texts.Aiming at the problem of cross-category in cross-border ethnic cultural data classification,this paper proposes a cross-border ethnic text classification method based on domain knowledge map,which uses cross-border ethnic cultural knowledge map to expand the semantics of cross-border ethnic entities in the text,and enhances the category semantic features of the text through the category features of entities in the knowledge map.In addition,by using the mask self-attention mechanism,the local features and global features of sentences in the text are obtained by extracting features at word level and sentence level,respectively.The experiments show that the F1 value of this method in cross-border ethnic culture data set is improved by 11.9%compared with the baseline model.

关 键 词：跨境民族文化 文本分类 领域知识图谱 实体语义扩展 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]