Methods of processing textual information in entity alignment algorithms

Entity alignment algorithms aim to find equivalent entities in cross-lingual knowledge graphs, which is important for the task of obtaining information about real-world objects. Recently, several studies have been conducted on entity alignment algorithms on various datasets. Algorithms using information about entity names have shown a wide range of results. In this paper, we have conducted a study of this phenomenon. Work has been done to improve the quality of matching cross-language entity names in vector space. Also, experiments with the modern models of processing natural languages have been carried out. The information obtained has led to a significant increase in the accuracy of entity alignment on the English-Russian dataset.