Collaborative iteration joint entity alignment method and device based on translation model

Abstract

The invention discloses a collaborative iteration joint entity alignment method and device based on a translation model. The method comprises the following steps of processing a plurality of knowledgemaps through a data preprocessing module, carrying out statistics on the data of entities, relationships, attributes, attribute values, relationship triads and attribute triads in the plurality of knowledge maps, and generating an entity alignment seed set according to a preset frequency selection method; generating a relation triple joint embedding space through a relation triple joint embeddingmodule; generating an attribute triple joint embedding space through an attribute triple joint embedding module; selecting a pair of entities with the minimum semantic distance through a collaborative iterative training module to form an entity pair, and adding the entity pair meeting a preset distance threshold value to the entity alignment seed set to update the entity alignment seed set; and carrying out iteration until the entity pairs satisfying a preset distance threshold do not exist. The method can solve the problem of joint fusion of the plurality of knowledge maps and the problem ofthe structure heterogeneous knowledge map and the cross-language knowledge map entity alignment.

Description

technical field [0001] The present invention relates to the technical field of data processing, in particular to a translation model-based collaborative iterative joint entity alignment method and device. Background technique [0002] Large-scale knowledge graphs are becoming more and more important as the data basis supporting various complex applications. They have important applications in search, analysis, reasoning, recommendation, decision-making, and question answering. They are widely used in search engines, intelligent assistants, Applications such as translation systems, question answering systems, and intelligence analysis cover finance, education, medical care, transportation, commerce, entertainment, government affairs, and security. In this context, the correctness and completeness of the knowledge graph is particularly important. However, due to the diversity of knowledge sources and the heterogeneity of knowledge structures in different knowledge graphs, the...

AI Technical Summary

Problems solved by technology

However, due to the diversity of knowledge sources and the heterogeneity of knowledge structures in different knowledge graphs, the existing knowledge graphs often have the following problems: (1) There are many knowledge duplications or supplements between different knowledge graphs; (2) A single knowledge graph The knowledge quality is low, the information coverage is low, and the knowledge description is incomplete; the knowledge heterogeneity between different knowledge graphs is strong, which is not conducive to data sharing and integration; (3) many domain knowledge graphs are only for a small area in the field. class, cannot support the upper application system

It can be seen from the above that the existing entity alignment methods have the following problems: (1) the internal information of the knowledge graph cannot be fully utilized, there is a problem of data discarding, and there is a lack of practicability; (2) most methods convert the fusion problem of multiple knowledge graphs into For the problem of pairwise knowledge graph fusion, the calculation cost is too high, and the problem of joint fusion of multiple knowledge graphs has not been solved yet; (3) The method based on similarity calculation cannot solve the problems of structural heterogeneity and cross-lingual entity alignment. not universal

Images