Annotatr - scientific paper commenting done right

Hi, I'm one of the developers of annotatr.<p>First of all, it's great to see people having a look at it and offering encouragement and constructive feedback. Thanks.<p>Annotatr was more a proof of concept and fun hack than anything else. While Bosco and I did want to offer an alternative to existing commenting systems for scientific papers (Nature, PLoS, CiteULike, etc.) we also realised that a big difficulty is getting continued engagement.<p>Some time after we put it up, I adapted annotatr and turned it into mldiscuss, a site for commenting on papers from ICML 2010:<p><a href="http://mldiscuss.appspot.com/" rel="nofollow">http://mldiscuss.appspot.com/</a><p>Before, during, and after the conference this site had quite a bit of activity but has since died off. I guess the lesson here is that a narrow focus (a single conference) works better than a broad one (scientific papers) -- at least in the short term. I'll likely extend mldiscuss for this year's ICML to see if it can pick up a bit more momentum.

The code is on GitHub: <a href="https://github.com/mreid/annotatr" rel="nofollow">https://github.com/mreid/annotatr</a><p>I remember seeing it last year and it doesn't look to have been worked on recently. A great proof of concept, but it'll be hard to sustain development motivation: there aren't good opportunities for making money, and this type of work unfortunately doesn't generate a lot of recognition amongst scientists.

I should clarify that I am only a user - the site was made by <a href="http://mark.reid.name" rel="nofollow">http://mark.reid.name</a> and <a href="http://boscoh.com" rel="nofollow">http://boscoh.com</a>

There's a persistent cross-site script vulnerability on the front page.

ah, not quite done right.<p>The search interface really needs to be more like google scholar or web of knowledge (who really wants to type in "author:"einstein a" &#38;&#38; year:1905"). Narrowing by field/discipline is a must, esp. for some keywords.<p>Also, just discussing abstracts is pretty lame. The typeface/links to citulike and journal are too small. There's no info on how many times it was cited, or what papers it cited.<p>Not to complain, but I'll wait for the next guy. This won't work.

quick wishlist:<p>- interactive refinement of results as I type<p>- ability to restrict results to only papers that have been commented upon<p>- ability to restrict visibility of my comments to friends/colleagues<p>- ability to get notified if a particular abstract gets a threshold number of comments<p>As it stands, this is not a minimum viable product I will return to (though perhaps it would be if everyone else in my field used it). Nice try though, hopefully it will improve with iteration.

There are loads of duplicate papers. Searching, for example. for "latent dirichlet allocation" shows the same JMLR paper about four times in the top results.

Sorry guys, I gave you an exception while searching for Polish theses:<p>UnicodeEncodeError: 'ascii' codec can't encode character u'\u0144' in position 14: ordinal not in range(128)<p>EDIT: Also, after navigating to: <a href="http://annotatr.appspot.com/citeulike/article/57175" rel="nofollow">http://annotatr.appspot.com/citeulike/article/57175</a>, I've got: DownloadError: ApplicationError: 5

This would work a lot better if there were social integration with your friends and colleagues - maybe scraped from Facebook?

There seems to be a small bug, when I try to search for an author with a german umlaut in the name (for example "Kühn Eva") I get an error ("UnicodeEncodeError: 'ascii' codec can't encode character u'\xfc' in position 9: ordinal not in range(128)"), however searching after "Kuhn Eva" works.

Searching for nothing causes an error with stack trace:<p><a href="http://annotatr.appspot.com/search/all?q=" rel="nofollow">http://annotatr.appspot.com/search/all?q=</a>