arXiv now has a recommender system, provided by CORE

arXiv — the preprint repository — and CORE — no, not the Australian conference and journal ranking but the “world’s largest collection of open access research papers” — announced a partnership. This partnership allows arXiv to provide research-paper recommendations to its users, provided by the CORE recommendation API.

arXiv readers now have a faster way to find articles relevant to their interests. From an article abstract page, readers can simply activate the CORE Recommender to find additional open access research on similar topics.

The Recommender, part of the arXivLabs toolset, was developed by CORE, a global aggregator of open access scientific content, which provides access to millions of full texts. CORE’s mission is to aggregate all open access research outputs from repositories and journals worldwide and make them available to the public. In this way, CORE facilitates free unrestricted access to research for all.

https://blog.arxiv.org/2020/10/15/core-recommender-now-supports-article-discovery-on-arxiv/

While the news was announced a few days ago, I believe I saw the recommender-system already a few weeks ago on arXiv.

We had partnered with CORE and its founder Petr Knoth ourselves about two years ago. For our living lab in our recommendation API Mr. DLib, we compared the recommendations of CORE and Mr. DLib in JabRef, a reference management software. Click-Through Rates on recommendations in JabRef were around 0.2% — for both, our own recommendation API Mr. DLib and CORE’s API.

Hopefully, CORE and arXiv will release their numbers, too. It would be interesting to see how arXiv and CORE compare to JabRef and Mr. DLib (which nowadays achieve click-through rates of around 0.5% in JabRef).