Anyone doing AI?

N. Veenstra

unread,

Dec 4, 2025, 2:02:51 PM12/4/25

to ROR Technical Forum

Hi all, I'm working on fixing our external organization master list in Pure using ROR, OpenAlex and agentic AI. Trying to bring down 180K organisations to about 15K. So far I can fix 50% by bridging OpenAlex and ROR with Pure, and for the remaining 50% I'm having AI agents analyse what we have.

This results in some amazing stuff, one of which being that I now have a mapping between dozens of wording variants into a ror org, and a mapping between various org id's (most notably scopus -> ror). I'd really like to build an agent cluster that employs this logic 24/7, or fine tune a model to do affiliation cleaning. Is anyone else working on this?

Best,

Nick

ai test.png

pure_alternatives.png

Amanda French

unread,

Dec 8, 2025, 11:12:22 AM12/8/25

to ROR Technical Forum, nick.v...@rug.nl

Hi Nick,

That sounds like terrific work! We of course would be very interested to hear of anyone else who is working on this or wants to do this, specifically with AI. I don't know if anyone else on the list will respond, but I've found that searching GitHub is a good way to find people using ROR in various kinds of affiliation string matching tools. There are things like Allen AI's Semantic Scholar Affiliations Linker, but as far as I know that uses older Named Entity Recognition techniques. Hope you also saw our recent announcement of a new affiliation matching strategy in the ROR API -- there's a nice demo by Adam embedded in the post. https://doi.org/10.71938/zz90-g810

Cheers,

Amanda

ROR Technical Community Manager

N. Veenstra

unread,

Dec 9, 2025, 8:33:59 AM12/9/25

to ROR Technical Forum, Amanda French

Thanks Amanda, I will look around on GitHub. Once I have cleaned up the database I will have most of our organanizations mapped from Scopus ID to ROR ID as well, which might be of interest to store in the ROR database. Other (Pure) users can then benefit from a faster ROR implementation in their own database, it would be awesome if we could combine data from similar cleanups.