基于图计算的论文审稿自动推荐系统  

Recommendation system for paper reviewing based on graph computing

作　　者：谢玮[1,2] 沈一[1,2] 马永征[1] 

机构地区：[1]中国科学院计算机网络信息中心,北京100190 [2]中国科学院大学,北京100049

出　　处：《计算机应用研究》2016年第3期798-801,共4页Application Research of Computers

基　　金：中国科学院计算机网络信息中心"一三五"重点培育课题(CNIC_PY_1407)

摘　　要：针对传统审稿方式所存在的问题,设计了基于TextRank图算法思想的论文推荐系统,以实现论文审稿分配过程的自动化。系统通过加入词与词之间的影响力计算以及多文档集中逆文档频率IDF,实现关键词抽取部分,并使用基于余弦向量值的计算对抽取出的关键词向量进行相似度匹配,最后计算审稿人在各研究领域的影响力,实现论文的推荐。采用了综合考察准确率、召回率的F值作为评测指标,验证了该方法的有效性。在实际使用环境中,该系统具有较高的准确性与可靠性。Due to the problems of traditional paper reviewing method,this paper proposed a paper recommendation system on the basis of a graph algorithm called,TextRank algorithm to achieve the automation of the distribution in paper reviewing process. The system firstly extracted keywords from text by adding the computation of the influences among terms and the inverse document frequency. Then it made papers match reviewers＇ interests through computing the cosine similarity of each vector. Finally,it computed the influence of reviewers in each field in order to recommend accurately. When verifying the validity of the recommendation algorithm,the article used precision,recall and F-measure as evaluation indicator. As a result,the system is proved highly effective and reliable.

关 键 词：论文推荐 推荐系统 审稿分配 关键词抽取 相似度计算 图计算 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]