Knowledge Base Completion With Analogical Inference on Context Graphs

International audienceKnowledge base completion refers to the task of adding new, missing, links between entities. In this work we are interested in the problem of knowledge Graph (KG) incompleteness in general purpose knowledge bases like DBpedia and Wikidata. We propose an approach for discovering implicit triples using observed ones in the incomplete graph leveraging analogy structures deducted from a KG embedding model. We use a language modelling approach where semantic regularities between words are preserved which we adapt to entities and relations. We consider excerpts from large input graphs as a reduced and meaningful context for a set of entities of a given domain. The first results show that analogical inferences in the projected vector space is relevant to a link prediction task