摘要
对某个领域或问题进行学术调研是科研工作的基本需求,然而随着越来越多的科研人员投身研究,大量的学术成果不断涌现,信息过载使得快速有效的调研工作变得越发困难.文中旨在提出一种自动学术调研框架,基于用户给定的关键词查询推荐最值得调研的论文及作者,以辅助科研人员高效完成调研任务.面向某个领域或问题最值得调研的论文和作者,需要具备显著的权威度且能覆盖该领域或问题的不同方面.因此,文中提出了一种面向权威度和多样性的两阶段排序模型:首先引入了MutualRank模型,同时考虑论文及作者信息以更好地建模他们的权威度;接着利用PDRank模型融合权威度和差异性两个因素对论文和作者排序,最终得到权威度高、覆盖面广的调研结果.通过实验作者证明了MutualRank对于权威度的学习效果优于传统的PageRank,同时基于两阶段排序模型得到的调研结果也优于已有的基准方法.
Literature survey of domains or topics is the foundation of scientific research. Along with more and more researchers devoting themselves to their work, plenty of academic achievements come out continuously, which brings more difficulties to effective and efficient surveys.This paper aims at developing an automatic literature survey framework to help researchers survey effectively. This framework recommends papers and authors which are most worthwhile surveyed based on the keywords given by the user. These recommended papers and authors must be prestigious and cover different aspects of the domain or problem. This paper proposes a two-phase ranking model by simultaneously exploring prestige and diversity. Firstly we introduce MutualRank to learn the prestige of the papers as well as the authors by leveraging the two heterogeneous types of information. We then rank the authors and papers by using PDRank model which combines the prestige and diversity. Finally, we provide users with recommended survey results with high prestige and diversity. Experiments show that MutualRank is better than PageRank on modeling prestige, and the superior to the existing baseline methods. survey results based on two-phase ranking model is
出处
《计算机学报》
EI
CSCD
北大核心
2015年第2期365-373,共9页
Chinese Journal of Computers
基金
国家自然科学基金(2013CB329601
61100175)
国家"二四二"信息安全计划(2011F45
2012G129)
国家科技支撑计划(2012BAH39B04
2012BAH39B02)
欧盟第七研发框架计划(FP7-PIRSES-318939)资助~~
