Welcome to the 20th edition (2017.1.0)
This is first 2017 edition of the Ranking Web of Repositories, that it is published since 2008. The current edition has been updated with data collected during January 2017. Next edition will be published at the end of July 2017.
For contact us, please write directly to the editor, the email address is [email protected]. You can check his experience and expertise about these topics accessing the scientific papers he authored as described in his Google Scholar Citations profile.
The criteria for inclusion are very simple: The repositories should have their own web domain or subdomain and include at least peer-reviewed papers to be considered (services that contain only archives, databanks or learning objects are not ranked. Individual journals are excluded too).
Although there is a paper describing the Ranking, Aguillo, I.F., Ortega, J.L., Fernández, M., Utrilla, A.M. (2010). Indicators for a webometric Ranking of Open Access Repositories. Scientometrics, 82 (3): 477-486. (You can get the author's edition in Open Access), there are some IMPORTANT CHANGES in the methodology. The indicators and weighting are described below::
OA Repositories are intended to provide full text documents, not bibliographic records or long abstracts. We use Google for counting the total number of pages provided, including all the rich files, like those in pdf format. However it is impossible at the moment to identify the full papers, so our best strategy is to exclude CRIS (and specially PURE installations) from the Ranking. We believe that current generation of CRIS are not favoring the deposit of full text documents
Our measure of quality is compiled from a virtual "referendum" where the opinion about the repository contents is extrapolated from the external inlinks (citations?) they received. Several colleagues provided evidence that top sources of links can be distorting the correct ranking of many repositories, so specific measures were warranted. We decided to change the methodology for obtaining the visibility indicator, using the total number of Referring Subnets from our link info providers (Majestic & ahrefs).
Obviously the horrendous practice of recommending the use of handles or external pURLs for citing the items is strongly penalized.
RICH FILES to be changed to SOCIAL (10%)
Frequently we receive requests for suggestions to improving the performance in the Ranking. We think repositories are deposits for archiving documents but a tool for promoting these items, and today the best, more universal and visible way is to use the social networks of the so-called Web 2.0. We are now counting the mentions in these tools: Academia, Bibsonomy, CiteUlike, CrossRef, Datadryad, Facebook, Figshare, Google+, GitHub, Instagram, LinkedIn, Pinterest, Reddit, RenRen, ResearchGate, Scribd, SlideShare, Tumblr, Twitter, Vimeo, VKontakte, Weibo, Wikipedia (all languages), Wikipedia English, Wikia, Wikimedia, YouTube, Zenodo. The results are normalized for each one and then combined (mean).
Perhaps a bit surprisingly Google Scholar is not indexing all the repositories. They mention formal reasons including difficulties to access all the contents or problems parsing the elements of the items into bibliographic files. They provide suggestions regarding these issues, but we feel quality matters too. A few webmasters are adding low quality material, splitting documents in many different files or adding almost empty records. For this indicator we count all the files in the Google Scholar database, excluding citations and patents.
Published figures are RANKS (lower is better), intended for showing individual performances, but they are not the raw values used in the calculations.
Three different rankings
We offer three different rankings. The main one consists of ALL the subject (like Arxiv or RepEc) plus the institutional repositories, that are ranked by normalization against the best (maximum) value. The second ranking is built from the subset of only the institutional repositories. Although the raw values used are the same, the procedure is done against different maximums, so the ranks obtained are independent from the ones provide by the global ranking.
Finally, there are a number of repositories that are very difficult to classify, so they are completely excluded from the main ranking. We are working on a better solution, but for now they are ranked separately as Portals. Please, consider sending to us your portals of academic journals to be included in this section.
Policy change regarding duplicate domains
As in previous editions, our policy regarding the repositories with two or more web domains or sub-domains, is to collect data and rank the redirected domain even if the older domain obtain better metrics. These repositories are marked (1) before the name.
In a few cases we have detected bad practices regarding link farms (paying money for receiving links from third parties or faking those third parties). In the past we simply excluded those repositories without further explanation, but from this edition we changed the policy. Their visibility indicator is set to zero and the rank appears in the lists as 99999. We do not answer messages about these issues.
ResearchGate, Academia and Mendeley
The three cited services are ranked under the Portals. It is not a common situation, but sometimes authors are depositing their papers in these websites instead of their own institutional repositories. The limited added value provided by the current generation of repository software and the inability of library managers and repository webmasters in improving profiling and other customization options are leading to this unfortunate emerging trend.
For this reason, we are showing the great performance of these "competitors" for increasing awareness: It is badly needed to improve repositories with a strong end-user (authors) orientation.
After recent purchase of SSRN by Elsevier, we moved this repository to the Portals section.