基于图注意力网络的小样本知识图谱补全

Graph attention network for few-shot knowledge graph completion

提出了一种基于图注意力网络(graph attention network,GAT)的小样本知识图谱补全方法.该方法通过图注意力网络的注意力机制赋予邻居不同的权重,生成更强大的特征表示,通过匹配网络匹配查询集与参考集,选择相似性度量分数最高的候选实体作为补全后的尾实体.实验结果表明,图注意力网络模型对小样本知识图谱中缺失的链接能够进行有效的预测.

In this paper,we propose a few-shot knowledge graph completion method based on Graph Attention Network(GAT),which gives different weights to neighbors through the attention mechanism of GAT to generate a more powerful feature representation.By matching the query set and reference set through the matching network,the candidate entity with the highest similarity score is selected as the completed tail entity.The experimental results show that the graph attention network can effectively predict the missing links in the few-shot knowledge graph.