基于多头自注意力模型的本体匹配方法  

Ontology Matching Method Based on Multi-Head Self-Attention Model

作　　者：吴楠 唐雪明 WU Nan;TANG Xueming(School of Computer and Information Engineering,Nanning Normal University,Nanning 530199,China;School of Physics and Electronics,Nanning Normal University,Nanning 530199,China)

机构地区：[1]南宁师范大学计算机与信息工程学院,广西南宁530199 [2]南宁师范大学物理与电子学院,广西南宁530199

出　　处：《无线电通信技术》2023年第6期1081-1087,共7页Radio Communications Technology

基　　金：广西研究生教育创新计划项目(YCSW2023437)~~。

摘　　要：随着语义网的发展,本体数量不断增加,本体间的语义关系变得越来越复杂。因此,引入OWL2Vec*方法获取本体的语义嵌入表示。通常,匹配的类或属性具有相似的结构,因此利用了字符级和结构级的相似性度量。为高效融合多种相似度值,提出基于多头自注意力模型的本体匹配方法(Ontology Matching Method Based on Multi-Head Self-Attention, OM-MHSA)自主学习各相似度方法对匹配结果的贡献值。在国际本体对齐评测组织(Ontology Alignment Evaluation Initiative, OAEI)提供的Conference数据集上进行实验,结果表明,相对LSMatch和KGMatcher+方法,提出的模型准确率(Precision)提升了6%,召回率(Recall)和F1值(F1-measure)超过了ALIOn、TOMATO和Matcha等方法。可见,提出的模型能够提升匹配结果的效率。With the development of the Semantic Web,the number of ontologies continues to increase,which leads to the semantic relationships between ontologies becoming increasingly complicated.Consequently,an OWL2Vec*approach is introduced to obtain semantic embedding representations of ontologies.Typically,matching classes or properties have similar structures,thus utilizing character-level and structural-level similarity metrics.In addition,to integrate multiple similarity values efficiently,the Ontology Matching Method Based on the Multi-Head Self-Attention(OM-MHSA)Model is proposed to independently learn the contribution value of each similarity approach to the matching results.Experiments on the Conference dataset delivered by the Ontology Alignment Evaluation Initiative indicate that the proposed method enhances the Precision by 6%compared with LSMatch and KGMatcher+methods,and the Recall and F1-measure exceeds methods such as ALIOn,TOMATO,and Matcha.It can be noted that the proposed model can enhance the efficiency of matching results.

关 键 词：语义关系 OWL2Vec* 本体匹配 多头自注意力模型 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]