citeulike and it's kin do this in a social-network sort of way. (Here are users that posted similar articles). I agree, though - building up a text corpus from existing papers and using it to make recommendations could be a fun and useful project. - Chris Miller

i've been saying this for years -- academic publications should have (a) amazon-style recommendations, and (b) flickr-style "interestingness". - Joe Dunckley

This is a problem of data, not of technology. Amazon has millions of people with a clear clickstream through a website. We've got people with PDFs on their desktops. MESUR is a good start at addressing some of that, but it will still only help for high-traffic areas. - Richard Akerman

What about someone behind an academic firewall? As long as I'm not selling the service, would it be a breach of copyright to mine the journals that my school already pays for access to? - Chris Miller

Yeah, the Amazon system wouldn't work without all that underlying data. That said, shouldn't citeulike, Papers, etc just make this part of what they do? - Deepak Singh

I think http://www.Biowizard.com does something like that. - Berci Mesko, MD

@ChrisMiller Most journal licenses explicitly prohibit programmatic access to the journal article contents. That is, any crawl, harvest, or machine processing of the metadata or fulltext is a violation of the licensing terms. (I'm not defending this position, I'm just stating the current situation.) - Richard Akerman

Perhaps you could focus on the OA journals/repositories and their articles? That might increase the acceptance of OA while helping scholars retrieve relevant/significant research free of charge. - Wobbler

It's really annoying for me that I miss relevant articles because I don't go to pubmed regularly and repeat the same searches. More than just repeating a search I need also to search horizontally via author name or by following the citation tree. - Michael Barton

I agree with wobbler. If BMC + PLoS could create some kind of combined service across their literature, would this not provide more citations for their journals? More citations => higher impact factor => more authors ponying up open access fees to publish there. - Michael Barton

@Michael I just remembered (since I'm doing some PMC Canada planning this afternoon) that the next phase of UKPMC will involve text mining, they're partnered with EBI and NACTEM. That should provide some quite powerful results across the free (but otherwise not fully open) articles in UKPMC. - Richard Akerman