We are currently implementing a research-paper recommender system that shows recommendations from Mr. DLib in JabRef (see Integrating related articles from MrDlib into JabRef ). In the first version, this recommender system will be able to show a list of articles that are related to the currently selected article.
However, at the moment, Mr. DLib contains primarily documents from Sowiport, one of Mr. DLib’s pilot partners. Sowiport has around 10,000,000 documents but mostly from the social sciences, mostly German, and only few of them are accessible in full-text. I guess, this is not the ideal document corpus for the JabRef users.
Therefore, I would like to ask if you know of any document corpora that we could include in Mr. DLib and recommend to the JabRef users. My preferred choice - for the beginning - would be arXiv.org (some million documents, all available in full-text, mostly in the field of math and computer science). Any other ideas?
Based on the user survey in 2015
JabRef users are coming mainly from the natural sciences.
So arxiv is already a good start (covering large parts of physics and mathematics). Some other database for biology, computer sciences or engineering would be also great. The following overview might help you: http://libguides.mit.edu/apis
I was wondering if it is possible to integrate with the digital libraries of universities and institutes.
They already have tens or hundreds of partners from different disciplines and payed access. This expands the results to non-public articles and journals.
we would go first for the CORE dataset, which has some 20,000,000 open access records. the disadvantage is that it is updated only once a year but as a starting point i guess this will be a really nice resource for the JabRef users to receive recommendations from. (maybe as a second step we could integrate the arxiv data and update it more regularly)