A neural translating general hyperplane for knowledge graph embedding

摘要: Abstract For completing knowledge graph, many translation-based models, like that Trans(E and H) which embed a graph into continuous vector space encode relations as translation operations in space, have achieved better performance. However, most of them limitations expressing complex for graph. In this paper, we propose translation-neural based method NTransGH completion. combines mechanism modeling by generalized hyperplanes, neural network capturing more interactions between entities relations. We conduct experiment on two tasks link prediction triplet classification with datasets. Experimental results show has strong expression mapping properties relations, achieves significant consistent improvements over state-of-the-art embedding methods. This paper is an extension our previous works [1] .