Graph completeness method based on knowledge graph neighborhood structure

Abstract

The invention provides a knowledge graph completion technology based on a neighborhood structure, aiming at solving the problem that a triple lacks in a knowledge graph. According to the technology, based on information such as entity neighborhoods, relation neighborhoods and corresponding relations between entities of the knowledge graph, modeling is conducted on relation elements and entity elements of the knowledge graph. The method mainly comprises the following steps: (1) establishing a model based on a neighborhood structure of an entity in a map, and mapping entity elements into an entity vector space; (2) establishing a model based on a neighborhood structure of relation elements in the map, and mapping the relation into a relation vector space; and (3) mapping the entity representation into a corresponding relation space by adopting a relation mapping matrix, and establishing a triple association model. In order to more effectively train the model, the invention provides a negative sample sampling algorithm based on a neighborhood structure, performs joint training on entities and relations, and predicts an unknown triple based on a training result. The contribution of theinvention lies in providing an effective knowledge graph completeness technology based on a neighborhood structure.

Description

technical field [0001] The invention belongs to the technical field of knowledge graph representation and reasoning, and in particular relates to a graph completion method based on the knowledge graph neighborhood structure. Background technique [0002] With the rapid popularization of the Internet, the contents on the Internet are diversified and the organizational structure is loose, which makes people unable to obtain information and knowledge effectively and quickly. In 2012, Google proposed the knowledge map, which opened up a new situation for the knowledgeization of the Internet era. Nowadays, knowledge graph technology has been regarded as one of the key technologies, and it is widely used in the fields of intelligent question answering and personalized recommendation. [0003] The knowledge map completion technology is proposed to solve the problem that there are still a large number of triples missing in the existing knowledge map. Existing research methods main...

AI Technical Summary

Problems solved by technology

[0006] Among the above-mentioned types of models, the model based on the graph vector mainly independently models the triples of the knowledge graph, ignoring the interconnection between the triples and the overall structure of the knowledge graph itself; the model based on the graph structure usually Treating entities and relationships as graph nodes with the same status, ignoring that entities and relationships are two completely different elements in terms of grammatical status and natural attributes, and cannot effectively use the different neighborhood information of entities and relationships; based on The deep learning model has the disadvantages of high computational complexity and poor interpretability.

Images