A BIBLIOMETRIC OVERVIEW OF OPEN SCIENCE RESEARCH

Abstract

Scientific research exhibits a vast increase in terms of published works, as well as produced data, computer codes, and scientific methods. Accordingly, openness is a modern demand in science, and many efforts occur in this direction. The reproducibility of results, data sharing, transparency, and replication are emerging challenges in research, as by following “open science” principles, the research community, as well as society, obtains beneficial outcomes in the best possible way. In this work, we computationally analyse a large volume of papers related to open science. Particularly, we use the Scopus database and search for papers with the term “open science” in the title, abstract or keywords, within the past ten years. In total 4878 papers were identified, in the following categories: Article (N=2770), Conference paper (N=724), Review (N=646), Editorial (N=188), Note (N=176), Book chapter (N=119), Erratum (N=112), Conference review (N=54), Data paper (N=30), Letter (N=29), Book (N=16), and Short survey (N=14). We start with descriptive statistics of the keywor ds appearing in the papers, and continue with inter-items' associations, by computing the contingency matrix. Accordingly, we calculate the objects' potions on the bibliometric map, by using a novel multidimensional scaling algorithm. In order to verify the associations, we permute the contingency matrix to minimise its bandwidth and plot the resulting clusters of keywor ds. Finally, in order to evaluate the evolution of the terms in recent years, we compute the timeseries of the occurrence of keywords and select the ones exhibiting higher trends, based on linear as well as nonlinear model fit. We also compute the normalised timeseries, in order to subtract the overall increase in scientific output. We discuss significant conclusions on open science research, based on a rigorous computational procedure.