Cross-Language Entity Alignment Based on Dual-Relation Graph and Neighbor Entity Screening

Graph convolutional network-based methods have become mainstream for cross-language entity alignment. The graph network has multi-order characteristics that not only process data more conveniently but also reduce the interference of noise effectively. Although existing achieved good results task alignment, they often overlooked same names in real corpus, resulting an entity-matching result was ideal. Therefore, this study proposed a neighboring-entity-screening rule by combining name and attribute (NENA) to influence these issues. We used NENA-screening filter delete redundant equivalent entities construct dual-relation as auxiliary evidence scenarios when information may be insufficient.This adopted order embed knowledge graphs into unified vector space, then down-sampling method extract neighboring each entity, thus forming sub-graphs two graphs. embedded GCN, new input, we cross-graph-matching module finally achieve Our on DBP15K dataset showed our approach significantly improved overall alignment.On sub-dataset ZH-EN DBP15K, value Hits@1 1.38%, compared best mentioned paper, it useful construction completion open graph.