一种基于标题与摘要语义的学术论文推荐方法

An Academic Paper Recommendation Method Based on Title and Abstract Semantics

本文提出了一种基于标题与摘要语义关系的融合网络模型进行论文推荐。首先,在词级子网络中,探究论文中标题与摘要的语义关系,得到论文的句级特征表示。其次,将词级子网络获取的信息输入到句级融合子网络中,对用户阅读偏好进行建模,最终得到为科研人员推荐的学术论文列表。在CiteULike-a数据集上的实验结果表明,本文所提方法较其他传统推荐方法取得了更好的结果,验证了该方法的有效性。

This paper proposes a fusion network model based on the semantic relationship between title and abstract for paper recommendation. Firstly, in the word level sub network, the semantic relationship between the title and the abstract in the paper is explored, and the sentence level feature representation of the paper is obtained. Secondly, the information obtained from the word level sub network is input into the sentence level fusion sub network to model the user’s reading preference, and finally the list of academic papers recommended for researchers is obtained. The experimental results on Citeulike-a data set show that the, the proposed method achieves better results than other traditional recommendation methods, and verifies the effectiveness of the method.