What I did is to take the entire library of articles I’m using for this journal paper and auto-extract them into Mendeley. Not bad – it’s getting better, but there’s still not one that I can leave untouched. Here’s the process I’m going through…

Sometimes it’s finding the DOI, which is AWESOME and it’s pulling that data. But in medicine the DOI lookup contains far less data (and is far-less correctly proofed) than the PMID dataset. So I’ve got a browser window open to PubMed, I’m simply copy/pasting the DOI number into the PubMed search field, hitting enter, and it’s giving me the record every time. From that, I copy the PMID back to Mendeley, hit search, and we’re doing great so far, with a couple of minor annoyances (doesn’t copy DOI back into Mendeley, not getting as many fields as it could). Honestly, I think you need a button next to the PMID field in addition to lookup called “Lookup PMID via DOI.” Seriously. PMID >> DOI.

Other times, it hasn’t been able to find the DOI but is almost always getting the journal title correct. In this case, I’m just copying the journal title into PubMed and hitting search, and it’s giving me usually just one, sometimes up to 4 hits. But it’s easy to get the PMID from that list, which again, I just paste back into Mendeley.

Now here’s where it almost hurts to see potential missed. Once you have a PMID, you can pull some phenomenally useful stuff straight off the tubes. See what I mean by going to http://www.pubmed.com, just put “17492789″ in the search field and hit enter. You’ll get an article by Karlsson and Breall on stent thrombosis. But down at the bottom of the page, there’s a drop-down list called “Display.” Change that from “AbstractPlus” to MEDLINE. See how it separates 1st and second authors? MH stands for MESH terms – things metasites like OVID use as keywords (hint). It also gives you the correct abbreviation for the paper for use in citations/biblios/footnotes (the TA field) and (hugely important) the complete abstract (the AB field).

OK, now scroll down again to Display and change to XML – that’s for the import filter if you’re not already using that to get PubMed import data. It’s killing me that when I import via PMID, it doesn’t also fetch the DOI, btw. Please – use all this data. It’s going to be necessary down the road. If nothing else, have Mendeley fetch all the data with field codes then map the field codes to Mendeley in the program. That way, if you’re not using the AB field today, at least you can build it in as a feature tomorrow without me having to re-lookup all my PMIDs

Now here’s the one where it gets REALLY useful. One more time to Display and change to “related articles.” When I was talking earlier about suggesting papers, here’s an easy way to make it work. Get the related papers list for everything you have in a specific “playlist” of articles, rank them by how many times each appears, then color code the ones that are already in the user’s library (green), then color code all the ones that free full text is avail for, either via pubmed or the Mendeley social service (blue) then those that there is no full text avail (black or red). Holy cow! Now I know which papers are highly likely to be cited by my audience and demanded by my critics.

OK – enough with the thought exercise. I hope this helps! All my best to you guys… You have unlimited potential with this program. Keep working hard. This is the part of development that separates the Googles of the world from the Infoseeks and Lycoses. You’re going to be great.

If it’s helpful, I’ll keep giving my feedback. Just think of me fondly when your commercial version costs $500 and no researcher can live without it

I think the conclusion that I keep arriving at over and over again for myself is that EndNote is a product of necessary evil. I’m writing a manuscript for publication right now, and hate every minute of setting up the library in X2. But for Mendely to make its first inroads into acceptance, it has to help rapidly set up a library. EVERY OTHER FEATURE is entirely secondary to the ease with which people can get papers into the library in the first place. I just spent literally 3 hours looking up the PMIDs for all these papers I want to import, and there’s only 40 cites so far. It’s maddening.

That’s why I sound like a broken record urging you guys to focus first on combining algoriths for recognition with some sort of social networking with fingerprinting so that I only have to do this once for the rest of my life – less if someone else has identified the paper once before. Get that done, and polish the word plugin so that it can at least use NEJM citation style or NLM citation style and you’ll be golden from medicine. The rest is gravy.

Can’t wait to see the new betas. In the meantime, back to End-of-me-Note. Have a great weekend!

The only thing I would be cautious about with social networking is, well, first how rivalries are handled and real experts are given a rightfully earned, higher priority place (Google Knowl style), and secondly how to protect the system from infiltration (ala-buying Digg front page stories). I think the first step is to get the ‘underlings’ – students, science writers, other non-primary literature producers – integrated and then introduce paper rankings and more features allowing what amounts to critiquing as more influential mentor/PI/professor types begin to move in. The one thing the current paper review system has right is the need for experts who have earned their titles as experts. I mean, imagine if we threw open the doors to voting on papers submitted by users and the first ‘school’ to really bombard the system was some Creationist, Flat-Earth, Mystical, Loch Ness Monster institution… It would be a mess!

**Create a feature in Mendeley which allows you to “string” papers into a chronological sequence. This goes with next suggestion which is:

**Create an importance field where you can personally rank a paper from -2 (cheapens the literature) to +2 (important step) with a special flag for “landmark” article.

**Once you’ve got those two, social network the stuff. I’ve got three papers on statin therapy in heart attacks strung together in a chain. Mendeley suggests that others with similar chains have also had the 4S and WOSCOPS studies flagged as landmark importance.

If it someday becomes possible to abstract the bibliography data of papers,
**Allow a user to assemble a chain of X papers and find out what studies are frequently cited in each – these must then be the seminal papers in that field.

**Have a feature in Mendeley where it reproduces the bibliography of the paper you’re looking at in a pane, with every paper you’ve already got in your library hotlinked in one color, and every paper that you don’t have, but that the author has made available in another. One click and bam – you’re looking at the next link in the chain.