Article Preview
Top1. Introduction
In recent years, people have entered an era of information explosion, due to the rapid development of information technology. A large number of data is being generated all the time in daily life. How to use these data more effectively to facilitate our lives is an urgent problem in the field of current information science. Along this line, the recommendation system came into being and was applied in many aspects of life. From e-commerce platforms, search engines, social platforms to short video platforms, portal websites, mobile applications, etc. All of them have certain recommendation functions (such as user behavior prediction, user interest perception, etc.). Traditional recommendation systems can be roughly divided into two types: content-based recommendation and collaborative filtering-based recommendation. They provide recommendations based on the similarity of the content, the users’ interactive behavior, and so on. These two classic recommendation systems are widely used because of their better recommendation effects. However, facing the current massive amount of information, traditional recommendation methods generally suffer from data sparseness and cold start(Guo, Zhuang, Qin, Zhu, Xie, Xiong, & He, 2020). To solve these problems, many studies have introduced knowledge graphs into recommendation research, hoping to realize interpretable recommendations from the perspective of semantics.
As a kind of semantic network, knowledge graph contains rich information. A typical knowledge graph consists of nodes and edges. Nodes and edges are used to represent entities and the relationships between entities. Specifically, the knowledge graph is composed of many “subject-predicate-object” triples, which can be denoted as “(S, P, O)”. Since the knowledge graph contains rich semantic relations, it is an effective attempt to use it as the information source of the recommendation system for the interpretable recommendation. Meanwhile, as a link in the use and creation of knowledge, the academic field are filled with a large amount of knowledge. This paper takes scholars as the subject, analyzes the research fields of scholars and the relationship between different scholars, uses scholars’ knowledge to construct a knowledge graph, and recommends related scholars based on the knowledge graph, which has certain theoretical significance for grasping academic trends, scientific and technological frontiers, as well as the development of research work and the introduction of talents.
1.1 A Background Example
A real example of scholar recommendation based on knowledge graph is shown in Figure 1.
Figure1. A Real Example of Scholar Recommendation Based on Knowledge Graph
It can be seen from Figure 1 that when the user “User” clicks on the scholars “Tang, Y.” and “Jiang, Y.”, we can assume that “User” pays more attention to these two scholars. By observing the nodes connected with the two scholars, we can find that the two scholars have the same research fields “Data Science”, the title “Professor” and the unit “SCNU”. Therefore, based on the nodes closely connected to these three nodes, we can infer that “User” may be more interested in the scholar “Gao, M.”, because this scholar also has the same title “Professor” and the research fields “Data Science”. At the same time, it can be inferred that “User” is more interested in the scholar “Zhu, J.”. Because he has the same title “Professor”, the unit “SCNU”, and he also has the same research fields “Social Network” with one of the scholars “Tang, Y.”. In addition, the scholar “Li, J.” has the same unit “SCNU” with the two scholars, but only has the same research fields “Social Network” with one of the scholars “Tang, Y.”. Therefore, we can infer that “User” may be interested in “Li, J.”. Similarly, for the scholar “Yang, L.”, although he shares the title “Professor” with the two scholars, there is no other connection to indicate that it is related to the two scholars. Therefore, we infer that “User” is less likely to be interested in him. It can be seen that in the research of introducing the knowledge graph into the recommendation system, a more reasonable recommendation effect and certain semantic interpretability can be obtained.