Knowledge representation method and system, storage medium and computer equipment

Abstract

The invention is suitable for the technical field of knowledge representation, and provides a knowledge representation method and system, a storage medium and computer device.The knowledge representation method is based on an existing knowledge graph and comprises the steps that multiple pieces of software downloaded by a user are obtained, and labels corresponding to the multiple pieces of software are obtained; similar labels of all the labels are obtained; the labels and the similar labels form a triad, the triad is a first triad, and the first triad forms a first knowledge graph; fusing the existing knowledge graph with the first knowledge graph to form a second knowledge graph; the triad in the second knowledge graph is a second triad; sampling the second knowledge graph based on a breadth-first search algorithm BFS to obtain a sample triple; and performing representation learning on all the sample triads. Therefore, knowledge can be enriched, and generalization performance can be improved.

Description

technical field [0001] The present invention relates to the technical field of knowledge representation, in particular to a knowledge representation method, system, storage medium and computer equipment. Background technique [0002] In computer vision, natural language processing, recommendation systems, neural networks, and deep learning are more and more widely used. In the application process of deep learning, Embedding, which converts discrete variables into continuous vectors, has greatly helped the application of neural networks in various aspects. Among them, learning based on knowledge graph representation has received widespread attention and has been successfully implemented in some industrial scenarios. [0003] In real engineering applications, there are cases of manual labeling knowledge or lack of knowledge. The cost of manually labeling knowledge is high, and manpower is wasted on labeling data. If knowledge is lacking, it will lead to insufficient knowledg...

AI Technical Summary

Problems solved by technology

[0003] In real engineering applications, there are situations where manual labeling knowledge or knowledge is missing

The cost of manually labeling knowledge is high, and manpower is wasted on labeling data. If the knowledge is lacking, it will lead to insufficient knowledge expression and poor learning effect.

At the same time, most of the current knowledge graph representation learning is mainly based on directly connected triplets to construct training data, which has low generalization and is difficult to learn network structure representation.

[0004] In summary, there are obviously inconveniences and defects in the actual use of the existing technology, so it is necessary to improve

Images