出版社：University of Sheffield, Department of Information Studies
摘要：Computer scientists and other researchers often make their programs freely available online. If this software makes a valuable contribution inside or outside of academia then its creators may want to demonstrate this with a suitable indicator, such as download counts. Download counts, citation counts, labels and licenses were extracted for programs that were both hosted in the Google Code software repository and cited in Scopus. Download counts were correlated with Web of Science citations, the distributions of both were compared and common software labels and licencing arrangements were identified. Although downloads correlate positively and significantly with Scopus citations, the correlation is weak (0.3) because some software has a large natural audience outside of academia. There is disagreement on the best licence to use for shared software, with no licence chosen by more than about a fifth of the projects. The most common language label was Java (20%) and, excluding generic computing terms, the most common topic labels were Google (5%), security (3%) and bioinformatics (3%). Download counts can give evidence of wider non-academic uses of software. However, software that is apparently not primarily designed for research but that is nevertheless cited by academics can also attract many downloads. Overall, download counts can be used as an indicator of academic value, but only if contextualised with the purpose of the program.
