Knowledge graph information representation learning method, system, equipment and terminal

Abstract

The invention belongs to the technical field of knowledge maps, and discloses a knowledge map information representation learning method, a system, equipment and terminal, and the knowledge map information representation learning method comprises the steps: carrying out the preprocessing according to a path constraint resource distribution method; calculating the reliability of all paths, and outputting the reliability to a training set and a test set; initializing the model and setting parameters; generating a triple according to an iterator, and randomly replacing head and tail entities; calculating a loss function of the triple according to the score function; calculating a loss function of an additional path according to the path reliability; performing parameter optimization by using an Adam method; and performing model verification by using entity prediction and relation prediction. According to the method, rich path information contained in the knowledge graph is considered, the modeling effect of entities and relationships is improved, the modeling of the relationships can be optimized by inputting vectors into a complex plane and using rotation to represent the vectors, and the method can be used for link prediction and recommendation systems.

Description

technical field [0001] The invention belongs to the technical field of knowledge graphs, and in particular relates to a knowledge graph information representation learning method, system, device and terminal. Background technique [0002] At present, Google proposed the concept of knowledge graph in 2012, which aims to represent unstructured or semi-structured information in the Internet as structured knowledge. With its powerful information processing capabilities and open organizational capabilities, knowledge graphs provide an opportunity for knowledge-based organizations and intelligent applications in the Internet age, and have been widely used in semantic intelligent search, personalized recommendation, and knowledge intelligent question answering. The knowledge map is developed from the semantic network. It is essentially a directed graph composed of entities and relationships. Each entity is used as a node of the directed graph, each relationship is used as an edge o...

AI Technical Summary

Problems solved by technology

Based on this directed graph representation, the related research and application of traditional knowledge graphs often need to be completed with the help of graph algorithms, facing the following two problems: - On the one hand, large-scale knowledge graphs often face the problem of data sparseness, It is difficult to achieve good results by using graph algorithms; on the other hand, graph algorithms often have high computational complexity and low computational efficiency, and cannot adapt to the application requirements of large-scale knowledge graphs.

[0009] (1) Traditional knowledge graphs need to be completed with the help of graph algorithms, but large-scale knowledge graphs often face the problem of data sparsity, and it is difficult to achieve good results using graph algorithms; at the same time, graph algorithms often have high computational complexity and low computational efficiency. Low, unable to meet the application requirements of large-scale knowledge maps

[0010] (2) Knowledge graphs based on symbols and logic are neither easy to handle nor sufficiently convergent for large-scale knowledge graphs; embedding models require high time complexity and large memory space, so it is difficult to Good results are shown on the knowledge map

[0011] (3) The distance-based model SE cannot accurately describe the connection between two entities because of the use of two independent matrices; the operation of the semantic matching energy model SME is more complicated; the complexity of the tensor-based model NTN is very high

[0012] (4) In the tensor-based model, when the scale of the knowledge graph continues to increase, the dimension of the tensor will increase, and the computational complexity will increase accordingly. Therefore, the tensor-based model is not suitable for large-scale knowledge graph representation learning. Can't show good effect

[0013] (5) The translation-based model TransE is too simple, and entities and relationships are represented by a single vector, which cannot accurately describe complex relationships such as reflexive, 1-N, N-1, and N-N

[0014] (6) At present, most methods regard triplets as a collection of independent triplets, and do not consider the association of triplets on the graph from the perspective of the graph, and most of the existing technologies are not enough to model the relationship Complete

[0015] The difficulty of solving the above problems and defects is: comprehensive consideration of each triplet will increase the time complexity and space complexity of model training, making the training time longer and occupying more space

Images