摘要
引文推荐旨在根据指定查询信息从海量数据中挖掘出与之最相关的若干文献,是一件有重要意义且极具挑战性的工作.引文推荐不但与文献的内容相关,文献间的引文关系、文献和出版社关系、文献和作者关系等,在引文推荐中也起重要作用.本文提出一种基于异质信息网络表示学习的引文推荐算法.首先,利用文献的内容信息,以及文献中的不同类型节点之间的相互关系构建异质信息网络;接着,对每个论文节点进行采样,对其先后进行元路径游走和随机游走,生成混合随机游走序列;最后,使用skip-gram模型获得节点的嵌入向量,计算相似性获得相应的文献推荐列表.在两个真实引文网络数据集上的实验结果表明,本文的方法在推荐效果上面优于已有的算法.
Citation recommendation aims to find potentially related papers by the relevant information of a paper.It is important and challenging to find related papers to cite from lots of papers.Lots of useful information can be used besides the content of the paper,for example,the citation relationship between the paper,the relationship between the paper and the publisher,the relationship between the paper and the author.This paper proposes a citation recommendation method based on heterogeneous information network representation learning.Firstly,the heterogeneous information network is built by using the content of the paper and the relationships between different types of nodes.Secondly,each node of the paper is sampled,and the meta-path walk is first performed,and then the random walk is used to sample the nodes to generate a mixed random walk sequence.Finally,the skim-gram model is used to obtain the representation of the node,and the recommendation list is generated based on the similarity of representation of the node.Experimental results on two real citation network datasets show that the proposed method can achieve better performance compared with the baselines of citation recommendation.
作者
段震
余豪
赵姝
陈洁
张燕平
DUAN Zhen;YU Hao;ZHAO Shu;CHEN Jie;ZHANG Yan-ping(Key Laboratory of Intelligent Computing and Signal Processing,Ministry of Education,Anhui University,Hefei 230601,China;School of Computer Science and Technology,Anhui University,Hefei 230601,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2021年第8期1591-1597,共7页
Journal of Chinese Computer Systems
基金
国家重点研发计划项目(2017YFB1401903)资助
国家自然科学基金项目(61876001,61602003,61673020,61702003)资助
安徽省自然科学基金项目(1708085QF156)资助
国家社科基金重大项目(18ZDA032)资助。
关键词
引文推荐
网络表示学习
网络嵌入
混合随机游走
异质信息网络
citation recommendation
network represents learning
network embedding
mixed random walk
heterogeneous information network
