A Survey of Knowledge Representation Learning Based on Structure and Semantics

Abstract

Knowledge representation methods have played an important role in the field of artificial intelligence especially in machine learning and deep learning. It converts useful information such as images, texts, and languages into low-dimensional and dense entity vectors, and provides NLP with better updated ideas and improves computational efficiency. In order to understand the current knowledge representation learning methods and status, this paper analyzes and categorizes the knowledge representation model based on structure and semantics, and finds that the knowledge represented by graph is easy to understand, but there are high complexity and long-tailed distribution, and semantic information of the relationship is difficult to obtain. Therefore, the semantic composition method of relation is adopted to solve this problem.