论文推荐系统的关键技术研究

Research on Key Technology of Paper Recommendation System

随着海量的研究论文出版发表,向研究人员推荐相关论文以满足他们的信息需求的论文推荐系统成为了一个重要的研究领域。论文相关度是论文推荐系统的核心,详细介绍了围绕这一核心的三类关键技术:引用关系分类技术、基于引用图的相关性度量技术和论文推荐算法,并实验对比了目前常用的五种相关性度量方法(共引、共联、CCIDF、HITS Vector-based和Katz距离)的推荐效果,由此提出用引用关系来量化论文之间的依赖关系,再结合Katz距离计算全局相关性这一改进意见.

With the tremendous amount of research publications, paper recommending system which recommends relevant papers to researchers to fulfill their information need becomes an important research area. This paper argues that paper relevance measurement is the core of paper recommending system. So three key technologies centering on this core are introduced in detail：citation relation classification, paper relevance measurement based on citation graph and paper recommendation algorithm.We evaluate five well-known approaches on a realworld publication data set and conduct an extensive comparison about them.At last, it is proposed to improve the global relevance of Katz by using reference relation to quantify the de-pendency between the papers.