A local adaptive knowledge graph optimization method based on transitive relationship

Abstract

The present invention provides a transfer relationship-based local adaptive knowledge graph optimization method, comprising: setting a training sample set; setting that any ri and ei initially belong to a certain distribution; performing normalization; forming a new training sample set; initializing a triple set as a null set; setting a correct triple corresponding to a wrong triple, replacing the correct triple with a head entity or tail entity of the wrong triple to form an error training sample set, and incorporating the error training sample set into the triple set; obtaining an edge parameter of an entity; obtaining an edge parameter of a relationship; calculating a parameter where the edge parameter changes with the entity and the relationship; obtaining a new loss function based on a transfer relationship; and performing determination and optimizing each entity or relationship vector by using a stochastic gradient descent (SGD) function. According to the present invention, the incompleteness of data can be made up, different potential semantics between the relationship and the entity can be better expressed, and the new knowledge map constructed after the optimization has higher accuracy.

Description

technical field [0001] The invention belongs to the field of knowledge management and information retrieval, and in particular relates to a local adaptive knowledge map optimization method based on transfer relations. Background technique [0002] Knowledge graph (Knowledge Graph) is a large-scale structured data collection, which is used to describe various entities and concepts in the real world, as well as the relationship between entities and entities, entities and concepts, and concepts and concepts, and expand the knowledge structure. , which in turn can be identified and analyzed by a computer. [0003] The knowledge graph consists of triplets composition( represent the head entity, relationship, and tail entity, respectively). In the current knowledge graph based on embedding translation, the main method is to embed entities and relations into the same low-dimensional space, or to embed entities and relations into different low-dimensional spaces. These methods...

AI Technical Summary

Problems solved by technology

However, if you continue to pass it like this, it must be considered wrong

[0006] From the above two examples, we can see that, on the one hand, knowledge is indeed transitive, so triples composed of entities and relations should also express this transitiveness; In the process, distortion occurs or some relations do not have the ability of knowledge transfer, making the newly obtained triples completely unrecognizable.

Images