基于模态相似性路径的统一多模态实体对齐  

Universal Multi-modal Entity Alignment via Iteratively Fusing Modality Similarity Paths

作　　者：朱柏霖 桂韬 张奇[1] ZHU Bolin;GUI Tao;ZHANG Qi(School of Computer Science,Fudan University,Shanghai 200443,China)

机构地区：[1]复旦大学计算机科学技术学院,上海200443

出　　处：《中文信息学报》2024年第6期34-44,共11页Journal of Chinese Information Processing

摘　　要：实体对齐(EA)的目标是从多个知识图谱(KG)中识别等价的实体对,并构建一个更全面、统一的知识图谱。大多数EA方法主要关注KG的结构模式,缺乏对多模态信息的探索。已有的一些多模态EA方法在这个领域做出了良好的尝试。但是,它们存在两个缺点:(1)针对不同模态信息采用复杂且不同的建模方式,导致模态建模不一致且建模低效;(2)由于EA中各模态间的异质性,模态融合效果往往不佳。为了解决这些挑战,该文提出了PathFusion,使用模态相似性路径作为信息载体,有效地合并来自不同模态的信息。在真实世界的数据集上的实验结果显示,与最先进的方法相比,PathFusion在Hits@1上提高了22.4%~28.9%,在MRR上提高了0.194~0.245,验证了PathFusion的优越性。The objective of Entity Alignment(EA)is to identify equivalent entity pairs from multiple Knowledge Graphs(KGs)and create a more comprehensive and unified KG.The majority of EA methods have primarily focused on the structural modality of KGs,lacking exploration of multi-modal information.A few multi-modal EA methods have made good attempts in this field.Still,they have two shortcomings:(1)inconsistent and inefficient modality modeling that designs complex and distinct models for each modality;(2)ineffective modality fusion due to the heterogeneous nature of modalities in EA.To tackle these challenges,we propose PathFusion,which effectively combines information from different modalities using the path as an information carrier.Experimental results on real-world datasets demonstrate the superiority of PathFusion over state-of-the-art methods,with 22.4%~28.9%absolute improvement on Hits@1,and 0.194~0.245 absolute improvement on MRR.

关 键 词：实体对齐 知识图谱 多模态学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]