Influence of Embeddings Construction Methods on Entity Alignment Approaches

Abstract

The problem of merging knowledge graphs (KG) presented in different languages is becoming more and more relevant. The main stage for its solution is the identification of equivalent entities and their descriptions. It is also known as the entity alignment problem. The recent research shows that existing approaches are not effective for all languages. This article presents the experiments aimed at improving the alignment of entities on an English-Russian dataset. The results obtained are considered from the point of view both of the whole graph and of individual types of entities. The influence of the number of relations and attributes on the accuracy of the algorithms is estimated.