dimensions = facets, you have some already, maybe also research methods? for social sciences and humanities, maybe theory? dunno. - Christina Pikas

I've always wondered why one would need so many different applications to do the same thing: handle scientific literature. One search engine to quickly find it, one or two other engines to find citations, publisher website with login or FF references wanted room to get the papers, Endnote, RefMan, Mendeley etc. for inserting refs in manuscript, CiteUlike, Connotea and Mendeley for collaborating/sharing/bookmarking, etc. - Björn Brembs

Ideally, there should be one place to find all papers, one place there to bookmark/share them and one way to take these bookmarked papers and add them to your manuscript as references. And of course, if you click on any reference in that paper, you get to the other paper. Mendeley for the first time has the potential to do all this with the power of crowdsourcing. Some of it would (sill) be illegal, but it could work. One ring, er, place to rule them all! Great post, Victor! - Björn Brembs

the finding part is actually tricky - different research areas/disciplines/domains have different information they need to determine if the article is relevant. So to have all the fields (like taxonomic classification, frequency range, or astronomical object number, subject age group) for every article in science doesn't work... you could look at lowest common denominator, and federated search does that, but it's really not where you want to go. - Christina Pikas

besides, some amount of competition is healthy. WoS didn't really make any significant changes for years until Scopus came along... However, interoperability is key as are things like apis and publishing and importing/exporting. - Christina Pikas

Recommender is tricky in science because there are many different domains. Is the most relevant paper to me as a computer scientist going to be the most relevant to you as a biologist? Amazon is making recommendations based on mass consumer behaviour - a single crowd. I'm not sure how well this will map to narrow scientific disciplines. I think you need something like "people in your field read this and also that". One thing I've been wondering about the b2x recommender is whether the user pool (presumably mostly undergrads doing course readings) will skew the recommendations. - Richard Akerman

Also, since FF doesn't have a way to connect "is-related-to", here's a related discussion thread: http://friendfeed.com/e... - Richard Akerman

right so following on from Richard - you might want related as in uses same algorithm, same species, same geographic location, ....these facets (geographic location) only appear in databases where they're relevant - Christina Pikas

@Christina And following on from my comment, my question continues to be, what standard facets are there that would be high value, but couldn't be machine-extracted. That is, what value can humans add in terms of facets (that aren't already done at creation time - for example many articles already include keywords)? (I can certainly see them adding value in terms of rankings, ratings, comments, folksonomy tagging etc.) - Richard Akerman

"Paper Pages" in Mendeley is a good step forward. Please consider integration with other services (e.g Connotea, CiteULike, Researchblogging.org, Nature Blogs) for these pages. - Martin Fenner

Martin: Yes, we'll do that! Richard: The concern for undergrads skewing collaborative filtering recommendation data and differing relevance for different disciplines was precisely the reason that we require academic position and discipline as part of our sign-up procedure - we felt that this was the minimum information that we needed (besides personal library contents) for making decent recommendations :-) - Victor / Mendeley Team

I also forgot to say that we can use Mendeley Document Groups and tags in the same way as the "explicit links" described in my blog post: If a Mendeley user puts two papers into the same Document Group, or assigns the same tag to them, then they should both be relevant to a common topic (i.e. intra-personal tags don't have the homonym problem of inter-personal tags, and could this be weighted more strongly)... - Victor / Mendeley Team

I'm looking forward to seeing the "Paper Pages" implementation - I think it has great potential. - Richard Akerman

I left my comments on your blog, Victor, but basically I think the "paper pages" idea is spot on. The kind of facets I was think about are things like "was trained by", "is colleague of", "is rebuttal of", "was influenced by". These are the kind of things that a text-mining approach won't be able to pull out from the data itself, but once you've got a set of annotations, you might just find some correlations that allow you to infer these properties. - Mr. Gunn

Just a plea for lots of rich feeds off of these paper pages - I want to monitor and mashup data coming from multiple sources about my and other interesting papers. - Cameron Neylon