AEVis: A Visualization Method to Facilitate Understanding Data and Entity Alignment Results

The cross-lingual Entity Alignment algorithms are designed to look for identical real-world objects in multilingual knowledge graphs. This problem occurs, for example, when searching for drugs, loT and other devices manufactured in different countries under different names. Embedding-based Entity Alignment algorithms have recently become extremely popular. The idea behind these algorithms is to obtain the descriptions of entities and relationships in the form of low-dimensional vectors so that the similarity or equivalence of entities and relationships is represented by the distance between their vectors. The advantage of the embedding-based EA approach is high scalability and low effort in preparing a test data set. Its main disadvantage is poor comprehensibility. This paper presents a visualization algorithm combining two visualizations: the conventional t-sne and graph-based. The set of options for the visualization tool developed makes it easier to understand not only the results of various entity alignment algorithms, but also the structure of the test data set.