Learning Attention-based Embeddings for Relation Prediction in nKnowledge Graphs

The recent proliferation of knowledge graphs n(KGs) coupled with incomplete or partial information, in the form of missing relations n(links) between entities, has fueled a lot of nresearch on knowledge base completion (also nknown as relation prediction). Several recent works suggest that convolutional neural nnetwork (CNN) based models generate richer nand more expressive feature embeddings and nhence also perform well on relation prediction. nHowever, we observe that these KG embeddings treat triples independently and thus fail nto cover the complex and hidden information nthat is inherently implicit in the local neighborhood surrounding a triple. To this effect, our npaper proposes a novel attention-based feature nembedding that captures both entity and relation features in any given entity’s neighborhood. Additionally, we also encapsulate relation clusters and multi-hop relations in our nmodel. Our empirical study offers insights ninto the efficacy of our attention-based model nand we show marked performance gains in ncomparison to state-of-the-art methods on all ndatasets.