Abstract
Objective: This study aims at devising novel bibliometric indicators to evaluate academic impact of research papers or researchers. Methods: Citation relation among articles (or other materials such as Web pages) is expressible as an oriented graph in which a node is an article and an edge is a citation link between an article and its reference. In addition, an oriented graph relating authors can be considered in which a node is a set of an author's articles (‘oeuvre’) and an edge is citation link(s) between two oeuvres. We devise and examine some impact indicators for articles and authors, using the simultaneous equations or the eigenvector for a graph matrix build from those oriented graphs. Results and Discussions: Two models for impact evaluation of articles are devised: (a) Distributed Citation Model (DCM) and (b) Knowledge Pool Model (KPM). The both models contrive to ensure the existence of the solution for the simultaneous equations or the eigenvector as impact indicators, taking into account a feature of articles that citations are necessarily made into one direction (from present to past). Next, ResearcherImpact (RI) is proposed as a model giving authors' impact. This uses the number of articles by an author cited by another author as an element of the graph matrix. The RI model is applied to impact evaluation of 15 principal authors in the bibliometric field. Conclusion: (1) The impact evaluation models DCM, KPM and RI proposed here give high impact scores to articles/authors contributing to (=frequently cited by) articles/authors of high impact (‘hubs’), compared to the simple citation count method. They are effective, therefore, for identification of a high impact community. (2) When applying RI, authors to be evaluated must be in a same research field and contemporary with each other, and also the matter of self citations and citations within a same affiliated institute or a research group should be noted.
