Authors:
Masanao Ochi
1
;
Masanori Shiro
2
;
Jun’ichiro Mori
1
and
Ichiro Sakata
1
Affiliations:
1
Department of Technology Management for Innovation, Graduate School of Engineering, The University of Tokyo, Hongo 7-3-1, Bunkyo, Tokyo, Japan
;
2
HIRI, National Institute of Advanced Industrial Science and Technology, Umezono 1-1-1, Tsukuba, Ibaraki, Japan
Keyword(s):
Citation Analysis, Scientific Impact, Graph Neural Network, BERT.
Abstract:
With the increasing digital publication of scientific literature and the fragmentation of research, it is becoming more and more difficult to find promising papers. Of course, we can examine the contents of a large number of papers, but it is easier to look at the references cited. Therefore, we want to know whether a paper is promising or not based only on its content and citation information. This paper proposes a method of extracting and clustering the content and citations of papers as distributed representations and comparing them using the same criteria. This method clarifies whether the future promising papers will be biased toward content or citations. We evaluated the proposed method by comparing the distribution of the papers that would become the top-cited papers three years later among the papers published in 2009. As a result, we found that the citation information is 39.9% easier to identify the papers that will be the top-cited papers in the future than the content inf
ormation. This analysis will provide a basis for developing more general models for early prediction of the impact of various scientific researches and trends in science and technology.
(More)