基于全文本分析的引文指标研究——以F1000推荐论文为例

Research on the Citation Indexes Based on Full-text Analysis:Taking the F1000 Articles as Examples

利用全文本分析法探讨更细粒度的引文指标计算方式,有利于更好地反映真实的引用情况,为引文指标的选择和改进提供参考。本文通过施引文献全文数据的获取、引用语境识别、引用信息识别、引文指标计算等步骤来讨论基于全文本分析的引文指标的获取和计算问题,再以F1000平台上的57篇论文为例,获取相关施引文献数据,基于引用强度、引用位置、引用情感等信息计算引文指标,对引用价值进行深入的分析,研究结果表明,学术论文在同一篇施引文献中存在多次被引,传统基于参考文献列表的被引次数计算方式不能真实反映学术论文的被引情况,对实证论文的引用多为不带明显情感倾向的中性引用,负面引用极少;实证论文被引超过半数发生在施引文献的结果位置,且在结果位置发生被引的比例越高,论文越受到领域专家的偏好。

Using full-text analysis to explore the calculation method of finer-grained citation indexes is conducive to better reflect the real citation situation,and provide reference for the selection and improvement of citation indexes.This paper will discuss the acquisition and calculation of citation indexes based on full-text analysis through the steps of citation full-text data acquisition,citation context recognition,citation information recognition and citation index calculation.Then taking 57 papers recommended on F1000 as examples to calculate the citation indexes based on citation frequency,citation location and citation emotion,the citation value is further analyzed.The results show that academic papers are cited many times in the same citing literature,and the traditional calculation method of citation based on references list could not truly reflect the citation situation of academic papers.Most of the citations to the empirical papers are neutral ones without obvious emotional tendencies,and few are negative ones.More than half of the citations occur in the result part of the citing literature,and the higher the proportion of citation occurs in the result part,the more preferred the paper is by domain experts.