[4Pin1] インタラクティブ(2)

[4Pin1-21] A Note on Similar Researcher Retrieval Considering Temporal Changes of Research Content and Affiliations

〇Hiroyuki Nishizawa¹, Marie Katsurai¹, Ikki Ohmukai², Hideaki Takeda² (1. Doshisha University, 2. National Institute of Informatics)

Keywords:researcher profiling, academic databases, research collaborator retrieval

As a research theme becomes more complicated, the range of knowledge necessary for research spreads wider. The collaborative research by multiple researchers is known to be effective for doing complicated research. Existing methods on research collaborator recommendation calculate the relevance between researchers using research content similarity and co-authorship relationships. This paper focuses on introducing new features, i.e., temporal changes of research content and affiliations, into researcher profiling for collaborator recommendation. In the proposed method, we use the titles and abstracts of academic papers as textual information for modeling research topics.
Specifically, calculating a topic vector of a researcher's publications in each year, we represent the researcher's interests using a series of topic vectors. We also obtain each researcher's affiliation information from the Database of Grants-in-Aid for Scientific Researcher, named KAKEN. Based on the resulting profiles consisting of temporal changes of research content and affiliations, we present a novel similarity measure of researchers. In experiments, we present topic extraction results from a subset of CiNii Articles and show an interface example of similar researcher retrieval based on the proposed method.