nodalpoint.org - Buggotea: Redundant links in Connotea - Comments

Thanks Ian, buggotea is almost fixed

Duncan — Wed, 28 Nov 2007 08:10:28 -0500

Thank you Ian for doing this. Connotea seems to perform better now. E.g. the BioPERL paper referred to in the original buggotea post
http://www.connotea.org/article/b0a2feab5e48494d3b03c39add264f22
has been bookmarked by "user tyrosine and 6 others" while
http://www.connotea.org/article/eff51b48308ba0bb333e598234c742aa
has been bookmarked by "user valine and 3 others". This is much better, because before I think they were all "user x and 0 others".

I think normalising URIs is a hard problem...

Buggotea, first pass fix

IanMulvany — Wed, 28 Nov 2007 07:34:10 -0500

Last Friday we released a major update to connotea with the aim of resolving Buggotea. Prior to this update the main database tables were something like this:

user
bookmark
user_bookmark
user_bookmark_comment
user_bookmark_details
user_bookmark_tag

and the unique key in the bookmark table was the hash of the url. This is where the entire problem with buggota stems from. All of the functionality of conontea was built on top of this structure. You could say that philosophically the structure was set up for people to share web pages that represented articles rather than articles represented by web pages.

We have changed the tables to the following structure:

user
article
bookmark
user_article (with preferred bookmark and citation pointers)
user_article_comment
user_article_details
user_article_tag

Where the article is an abstract entity more representative of the fact that people want to reference an abstract piece of work that can exist in many different locations. We now normalise bookmark entries via PMID and DOI into a single article object. It seems to be working so far for entries in connotea that have PMID and or DOI information.

This is a first step and for sure there are things that can be improved, support for normalisation based on other meta data, how the information about an article with multiple instances is displayed, concatenation of the best metadata from multiple authoritative sources. Its not rocket science but this is a good step towards making connotea more useful in the academic realm. Anyway, I thought you guy's might be interested. Let me know if you spot any odd behavior with the system.

Citeulike, Buggotea and DOI's

Duncan — Mon, 16 Jul 2007 07:25:21 -0400

Yes, citeulike doesn't suffer from buggotea, although it doesn't let you post DOIs as URIs e.g. in the form http://dx.doi.org/10.1101/gr.361602 it seems that citeulike generally spurns DOIs altogether?

I use this solution:

Madder — Mon, 08 Jan 2007 21:19:07 -0500

This problem bugged me badly, because my library was getting littered with numerous duplicate copies of the same Pubmed abstract. Because Pubmed URLs often contain session specific info like query_hl=10, the same abstract from 2 different sessions would get added twice to my library.

So I googled for "javascript tutorial" and then hacked the bookmarklet javascript to strip off the PMID from the URL and construct a new minimal URL in a consistent format to send to connotea. Another advantage of this bookmarklet is that in cases where a pubmed search returns only one abstract, the PMID is not in the URL at all (for instance, this search: search pubmed for term connotea). In this case, I can highlight the PMID on the page and then clicking the bookmarklet constructs the URL with the highlighted PMID and sends it to Connotea.

The bookmarklet javascript is here:

http://brahms.cpmc.columbia.edu/~suresh/connotea-bookmarklet.html

Obviously, everyone would have to use this solution to make their references consistent with the URL format I happened to use, but at least I dont end up enthusiastically adding 10 copies of the same *very important* abstract to my library myself.

HTH, S.

This is why I use citeUlike

Einat — Sun, 24 Dec 2006 08:10:02 -0500

It's good only for papers, but gets rid of the redundancy, and the abstract is in there.
Merry Christmas

Trumpton Riots

Duncan — Mon, 18 Dec 2006 03:55:52 -0500

Thanks to NPG for your rapid response. As for Half Man Half Biscuit, I suppose obscure 1980's english punk bands are a little "off topic" for a bioinformatics blog but I couldn't resist the reference!

HMHB

Neil — Sat, 16 Dec 2006 00:28:25 -0500

This is for sure the first reference to Half Man Half Biscuit on Nodal. I rate it highly purely for that reason :)

Us NPG pixies hear ya

stewc — Fri, 15 Dec 2006 09:58:14 -0500

This is something that we've been meaning to address on connotea.org for a while, but other more pressing concerns keep cropping up. The basic problem is that Connotea is keyed on URLs rather than identifiers or a combination of the two.

We could normalize the URLs, but I'm not sure that this is a good idea: we'd be making assumptions about what the user intended to bookmark in the first place (the paper? The abstract as it appears on PubMed? The free full text version on PubMedCentral? The paper as accessed through a proxy of some kind?).

What could / should be done is a fix to make the bookmarked by x other people snippet and URL info pages take shared identifiers into account. We'll definitely look at this in the new year, unless anybody fancies having a go with the open source release, in which case you rock and should feel free to email if you need any help.

Buggotea: Redundant links in Connotea

Duncan — Fri, 15 Dec 2006 07:19:26 -0500

Dear Santa, all I want for Christmas* is a better version of Connotea, please can you sort out it's duplicated redundant links? In my book this particular bug is “buggotea” number one. Here is the problem... [update: buggotea is partially fixed, see comments from Ian Mulvany below]