Feature of the week: 2009 in review

2009 brought many improvements and new features for BibSonomy but also interesting research activities. We briefly review this year before next week's post gives an outlook on 2010.

Tag Recommendations

As part of the ECML PKDD 2009 conference we organized the Discovery Challenge, where the participants could test their tag recommendation methods on a BibSonomy dataset. A particularly interesting part of the challenge was the online evaluation which allowed the researchers to evaluate their approaches in the running system and actually show their recommendations to our users. The underlying infrastructure was provided by our new tag recommendation framework which proved to be very useful. It allowed us to distribute the tag recommendation work over several machines located all over the world. E.g., the winner's recommender was running in Canada.

Research Projects

Two new projects centered around BibSonomy started this year: PUMA, which will improve academic publication management in cooperation with the University Library Kassel, and Info 2.0 (in German), which investigates chances and risks of the Web 2.0 with respect to informational self-determination in cooperation with the Institute for Public Law.

Plugins

We released three new plugins which better integrate BibSonomy with other tools. The JabRef plugin allows you to synchronize your publication references with the bibliography manager JabRef and the Typo3 extension integrates publication lists from BibSonomy into the content management system Typo3. Just released two weeks ago and ready for testing is the new Firefox add-on which better integrates BibSonomy into the Firefox web browser. We will introduce this add-on in one of the next FOTWs.

Personalization

You now see similar users in your sidebar on which you can click to surf their posts in a personalized ranking. Furthermore, you can follow users you find interesting to stay tuned on what they post.

Dumps of the Dataset

Since quite some time we offer a dataset of the BibSonomy database in form of an SQL dump for research purposes to interested people. A web page now describes the available dumps and how to get one. Newly, the dumps also contain the users' tag relations.

Development

In an ongoing effort to open the BibSonomy source code to the public, we released some of the core modules in a public Maven repository. E.g., now you have access to our screen scrapers, which allow you to extract publication metadata from more than 60 digital libraries. Most modules have a GPL or LGPL license.

Next week we will present our current activities and discuss the plans for 2010.

Stop thinking, start tagging - Tag Semantics emerge from Collaborative Verbosity

Maybe you've asked yourself from time to time "What are these BibSonomy developers doing the whole day?" Of course the first answer is simple - we develop BibSonomy :) - but apart from that, most of us are researchers, running experiment, discussing results, writing papers - and the latter is sometimes rewarded: Our work "Stop thinking, start tagging - Tag Semantics emerge from Collaborative Verbosity" got accepted at this year's WWW conference in Raleigh, USA!

As you can guess from the title, the paper is basically concerned with emergent semantics. This term is often used to describe semantic structures that "grow" in a bottom-up and uncontrolled manner within collaborative tagging systems. For the case of emergent tag semantics this means that despite people are free to choose arbitrary tags (which leads to typical language-related phenomenons like homonymy, polysemy, ..), one can successfully extract meaningful tag relations from the aggregated mass of tagged content. As an example, different people might use different tags to describe the web2.0 paradigm, possibly "web2.0", "web-2.0", "webtwo", "web20", "web.2.0", and many others. By using the appropriate tag relatedness measures, one can identify those cases and extract a semantic "concept" web2.0 which all these users are talking about.

Up to this point, there's nothing too new - the question we asked ourselves then was how the characteristics of individual users influence the quality of the learned semantic structures. One possibility is to distinguish users according to their tagging motivation into "Categorizers" and "Describers" - the first group uses a small and systematic vocabulary, wherby the latter uses a wealth of different keywords for annotation. Simply spoken, describers can be seen as the "verbose" users tagging with many keywords. So we splitted up the whole folksonomy dataset into several partitions containing different mixtures from categorizers and describers. And here is an interesting thing we found:

On the x-axis, you see the percentage of included users. The y-axis depicts the quality of the inferred semantic tag relations (measures by grounding against a thesaurus; as we used the JCN distance, smaller values indicate better quality). The green line depicts the semantic quality obtained from the full dataset. The interesting thing is now that already with 40% of the "talkative" describers, one can reach the semantic precision of the full dataset! The best quality is found for 70% of describers. So the claim that "mass matters" holds only partially - a crucial aspect seems to be from which kind of users the mass is composed. The collaborative verbosity of describers seems to have a positive effect on the emergent semantics. On a more general level, this exhibits a causal link between tagging pragmatics (how people tag) and tag semantics (what tags mean). If you're interested in further details, we'd be happy to discuss with you on WWW2010!

Discover Names

Today I'll introduce to you to one of our new projects at our research unit: The Nameling

The Nameling was developed in the need for finding a given name (which of course should be nice and enjoyable). It turned out that finding a nice name wasn't that difficult - but all names which we found pleasant, were also popular in our neighbourhood. Being busy at work with mining patterns in huge data sets, the idea for mining pleasant names was born.

Currently, the Nameling implements a similarity search for given names, based on co-occurrences in the English and German version of Wikipedia (as a hidden feature, you can also access similarities emerging from the French version). This allows you, for example, to browse for names which fit to a name you like. Whenever possible, we also extracted corresponding categories for names from Wikipedia, allowing you to browse through all names within a given category:

By clicking on the lovely butterfly, you can add names to your list of favourite names. You can also share your favourites with some of your friends, for collaboratively searching for your "best fit":

The Nameling gives you also access to further background information (at least a link to a corresponding Wikipedia article, if available). This comprises popularity rankings obtained from Wikipedia and Twitter, distribution of names over time as extracted from person data in Wikipedia and co-occurrences displayed as a graph whose nodes you can expand by a mouse click for navigating within the co-occurrence network.

Of course: We are a research unit and we want to go further. The Nameling currently implements a first and simple measure of similarity. We will make use of the Nameling's query logs to improve rankings and similarities. We will look for emerging patterns and correlations between user profile information and name preferences. The Nameling is still a very young project and surely needs improvements - but it already yields some astonishing results. Try it out yourself!

Happy browsing!
.folke