Entity alignment method based on combination graph structure information and text semantic model

Abstract

The invention discloses an entity alignment method based on combination of graph structure information and a text semantic model, which belongs to the technical field of computer networks, and comprises the following steps: extracting unstructured text data entity information, entity attribute information and entity relationship information, forming original triple data, and generating graph structure-based entity embedded representation; the method comprises the following steps of: extracting context related mechanism and name information of the entities in an original text as entity auxiliary description information; calculating whether the description information of the different entities is intersected or not; calculating different entity name editing distances and word2vec cosine similarity; calculating different entity name semantic similarity on the basis of a pre-training model; according to the method, the graph structure information, the character information and the semantic information are comprehensively utilized to judge the similarity between the entities, the entity graph structure information and the semantic information are fully utilized to carry out entity alignment, and the alignment accuracy is improved when common information is sparse.

Description

technical field [0001] The invention belongs to the technical field of computer networks, and specifically relates to an entity alignment method based on combining graph structure information and text semantic models. Background technique [0002] In recent years, the Internet has spread rapidly all over the world, and network services have begun to diversify. Different network information may provide the same or different information about the same thing at the same time. Therefore, for each entity in the knowledge base of heterogeneous data sources, find out Belonging to the same entity or concept in the real world Entity (concept) alignment is of great significance in many fields such as relational analysis, event reasoning, network security, etc. [0003] On the Internet, data describing the same entity usually appears in multiple places, for example, there are descriptions of the same entity in different encyclopedias. Therefore, when building a knowledge graph, especia...

AI Technical Summary

Problems solved by technology

[0007] Most of the entity alignment algorithms based on string similarity are carried out in the case of rich common attributes, and lack the use of structural information and text semantic information. Therefore, in the current entity alignment work of industry knowledge graphs, it is necessary to combine graph structure information and text semantic models. Entity Alignment Method for

Images