Identifying the representative works of scientists

Ranking the scientific publications is becoming increasingly important to uncover the quality of academic works and evaluate the performance of researchers. Though many ranking algorithms have been proposed and intensively studied, these methods only focus on the global ranking, i.e. ranking the general significance of papers. However, a long-neglected practical issue for both policy makers and scientists themselves is to identify the representative works of each scientist. This is not a trivial problem because the representative works of a research is not simply his most highly cited paper nor the paper published in top journals. It can happen that a physicist collaborates with an expert in biology and publishes a very highly cited paper in a top journal. This paper is a side interest of the physicist so of course shouldn’t be considered as his/her representative work. Therefore, it requires us to consider a personal ranking of the scientific publications instead of the global ranking of them. In this paper, we consider the representative works of a scientist as an important paper in his expert area. Accordingly, we propose a preferential diffusion process to generate personalized ranking of papers for each scientist and identify his/her representative work. We use the citation data from American Physical Society (APS) to validate our method. We find that the Nobel prize paper of the Nobel laureate is be ranked rather low in his/her personal ranking list if the citation count is used. With our preferential diffusion method, the ranking of the Nobel Prize papers can be significantly improved. Moreover, we test the robustness of our method. Results show that our method can highly rank the representative papers of scientists even partial publication data of them are available.