一种基于语言模型的科技论文推荐算法

Researchers usually need to consult relevant documents when they prepare for writing papers, and they have to identify the mass of papers. In order to improve the efficiency, this paper draws on the recommendation system which has been proved effective in business field and put forward a novel approach to help researchers. At first, this paper explains the concept of citation network, and then explains the relationship between citation and language model. Furthermore, this paper proposes the concept of citation sentence model. On this basis, a deep learning natural language processing word vector tool is used to train the citation sentence to get a model, which will be used to find recommend results that are most similar to target papers. The results of experiments show that the proposed approach achieves success when compared with traditional collaborative filtering methods and some others.