Researchers, their scientific publications and their research projects are often object of evaluation, from different points of view, for many different purposes. However, even if different metrics have been proposed in literature, they usually assume the co-authorship to be a proportional collaboration between the researchers, missing out their relationships and their change on time along the career. In this work, we propose an application that makes use of a novel metric for evaluating and comparing researchers by taking into account the co-operations among them and estimate their reciprocal dependence degree along time. This application can help comparing and ranking researchers based on his/her demonstrated independence, along his/her whole career, with respect to the surrounding research community.
Introduction
Bibliometric indicators are increasingly used to evaluate scientific careers based
on personal publication records. The simple number of papers published by an
author rather than the received citations are still common ways to capture both
the quantity and the impact of an author’s set of works. However, these methods
do not capture the actual contribution of a researcher within a research network.
In this respect, it has been much discussed whether co-authors should have all
the same value in quantifying the impact of a paper. In [7], for example, the
author first pointed out the problem of undeserved coauthorship. In [
In light of this, the pure information about the publication records of a researcher often results insufficient for a fair evaluation of scientific profiles because they do not take into account many factors, as the relationships between the authors and their relative scientific influences, which should be directly considered in the evaluation process.
In fact, especially when these measures are used for recruitment purposes, it is highly relevant to analyze the scientific dependencies among authors in order to estimate the capacity of an author to work and produce research outcomes without the people that assisted his or her work until that time.
A research collaboration can be indeed defined as a two-way process where individuals and/or organizations share learning, ideas and experiences to produce together scientific outcomes. Collaborations are necessary because of the evident difficulty for individual scientists to conduct several groundbreaking research on their own. For this, one of the key aspect of a successful researcher is the development of a large, active, network of collaborators that can help the researcher to bring new solutions and propose, continuously, novel ideas and approaches to the research community. On the other hand, evaluation of individuals needs a sort of inverse process with the primary goal of understanding the role of each researcher, and his/her specific impact on the research community, in this collaborative environment.
In light of this, following the work presented in [
With these goals in mind, based on the entire DBLP bibliographic database, in this paper we present a web platform (available at http://d-index.di.unito.it ), which allows the user to study the scientific profile of each researcher and analyze, through several dynamic visualization tools, the evolution of the impact of each collaboration on his/her scientific output. 2
Formalization of Scientific Collaborations in
Publication Networks
Based on the previous theoretical works proposed in [
Given two collaborating researchers (also called authors along the paper), ri, rj and their common scientific network Nrti,rj , defined as the set of researchers who collaborated with them, the autonomy of their collaboration atri,rj at time t is calculated as:
0 t ari,rj = Prk∈Nrti,rj 1 c(rk,Orti,rj ) x1 ! Px=1 if Nrti,rj = ∅ if Nrti,rj 6= ∅ where the function c(rk, Orti,rj ) returns the number of times a researcher rk coauthored a paper with both ri and rj at time t. The higher the autonomy the more independent the work of ri and rj is from their research environment. We then define the dependence value of ri on the collaboration with rj as dtri→rj as t dri→rj =
t pri,rj × pt ri ari,rj,Nrti + atrj,¬ri,Nrti
t ari,rj,Nrti + atrj,¬ri,Nrti + atri,¬rj,Nrti t , where p is a productivity score (number of published works) of a is the autonomy score. The dependence value dtrj→ri ranges from 0 to 1 ; in particular, dtri→rj ≈ 0 indicates that the dependence of ri on rj , at the time t, is negligible, while a dtri→rj ≈ 1 highlights the contrary.
Thus, given the complete set of dependence values, for each year and relative to each co-author, we calculate the researcher’s dependence trajectory, by calculating the standard deviation, along the time, of each dependence value, for each co-author, from the optimal attended value of 0 (which would mean a dependence score of 0; i.e., the production of the considered researcher is independent from the collaboration with the considered co-author). In a sense, we aim at evaluating the overall independence of a researcher from the surrounding community. More formally, given a researcher ri, we define his/her dependence −→ trajectory dri = {sdtri , sdtr+i1, · · · , sdtr+in}, where sdtri is calculated as sdtri = s Prk∈Nri (dtri→rk )2 |Nri | .
We can use these values to properly compare, and rank researchers with similar characteristics. More in detail, we provide a radar chart that can rank the independence performance of a considered researcher with respect to those who have i) similar career length, ii) similar number of publications, iii) similar number of co-authors. 3
Web Application and Real Case Scenario In this section, we introduce our application for analyzing, comparing and ranking scientific collaboration patterns of researchers. The web application is available at http://d-index.di.unito.it. As data input, we considered the DBLP data set1.
The proposed application permits to search for any author indexed by DBLP and to take a preview, through several features and visualizations, of his/her scientific profile and her/his collaboration history over time. The user can analyze the evolution over time of each scientific collaboration for a searched researcher. It is also possible to The system can visualize the evolution of the dependence of a researcher on the support of each co-author along the career. With this chart, it is also possible to select/deselect additional co-authors to make further analyses and comparisons. The application also provides a dynamic visualization chart 1 http://dblp.uni-trier.de/db (called “time-lapse”) which allows the user to focus on a specific time interval and/or a subset of co-authors.
Finally, the proposed tool tool also allows to compare and rank the overall independence of an author, along his/her whole career, with the whole research community. Please also notice that it is also possible to compare the considered researcher against others (even if they do not share the same time career). This visualization permits to focus on how much the entire production of a researcher can be considered dependent on the interactions with her/his local community.
The presented demo can be used to analyze each researcher in the entire DBLP community by also considering similar profiles (with parameters such as number of papers, number of co-authors, and length of career).