基于嵌入的实体对齐技术对比研究  

Comparative Research on Embedding-based Entity Alignment

作　　者：徐漫 刘爽 张云霄 朱晓敏 XU Man;LIU Shuang;ZHANG Yun-xiao;ZHU Xiao-min(School of Computer Science and Engineering, Dalian Minzu University, Dalian Liaoning 116650, China)

机构地区：[1]大连民族大学计算机科学与工程学院,辽宁大连116650

出　　处：《大连民族大学学报》2022年第3期264-270,275,共8页Journal of Dalian Minzu University

基　　金：国家自然科学基金资助项目(61876031)。

摘　　要：近年来,基于嵌入技术的实体对齐方法迅速增加,但是这些方法的相对性能模糊不清,缺乏完整的经验评估。将对最先进的实体对齐方法进行全面和详细的分析。提出一个适用于实体对齐方法的通用框架来捕捉技术的关键特征,并通过实体信息的不同方面来进行特征的分类处理。基于实体对齐任务中常见的数据集和评价指标,对十六种模型进行对比分析,说明融合不同方面信息可以取得更好的对齐效果。最后对实体对齐领域的未来发展进行探讨研究。Embedding-based entity alignment methods have increased rapidly in recent years,but the relative performance of these methods is ambiguous and a complete empirical evaluation is lacking.A comprehensive and detailed analysis of state-of-the-art entity alignment methods is presented in this paper.A general framework suitable for entity alignment methods is proposed to capture the key features of the technique and to classify the features through different aspects of entity information.Based on common datasets and evaluation indicators in entity alignment tasks,sixteen models are compared and analyzed,indicating that integrating different aspects of information can achieve better alignment results.Finally,the future development in the field of entity alignment is discussed.

关 键 词：知识库 知识融合 实体对齐 实体嵌入 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]