.png){kind=icon}
.png){kind=icon}
Return
Improving entity alignment via attribute and external knowledge filtering
.png){kind=icon}
PRE
AI
DOI：10.1007/s10489-022-03744-5.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The goal of entity alignment is to find the equivalent entity pairs in different Knowledge Graphs (KGs), which is a key step of KG fusion. Recent developments often take embedding-based methods, which mainly focus on embedding structure information (relationship triples) of KGs to align entities. However, attribute information (attribute triples) and external knowledge (some public knowledge bases) can also provide extremely valuable information but have not been well explored yet. In this paper, we propose a Selective Filtering Entity Alignment (SFEA) framework for improving entity alignment via attribute and external knowledge filtering. Our framework proposes two kinds of selective filtering mechanisms to filter the candidate set by using attribute information and external knowledge, so as to delete most of wrong entities in the candidate set. The framework mainly uses selective filtering mechanism and external knowledge for entity alignment. Experiments show that our framework achieves state-of-the-art on the dataset DBP15K compared to the existing similar methods of using structure and attribute information. Specifically, our framework outperforms the best state-of-the-art methods by 1.2%-4.9% in terms of Hits@1, and also achieves better or comparable performance when compared with other methods of incorporating multiple sources of information.
Keywords:
Entity alignment
Selective filtering
Knowledge graph
Neural networks
Journal
IF:
3.5
Papers:
7.5K
Citations:
1.7W