Our paper A hybrid approach to item recommendation in folksonomies has received the Best-Paper-Award at the Workshop on Exploiting Semantic Annotations in Information Retrieval (ESAIR 2009), co-located with this years WSDM conference in Barcelona.

The paper presents an extended version of the PLSA algorithm that combines collaborative filtering and tagging patterns found in folksonomies into a unified recommendation model. Experiments were done on our delicious corpus. We believe that folksonomies, and especially Social Bookmarking Services such as Delicious, provide a perfect soure for web resource recommendation.

If you want to get an idea what item recommendation based on folksonomies can look like see here.

Loosing against the Brainstomers, the later tournament winner, in the quarter finals, the DAINAMITE team arrived 6th at this years Robocup. 3 places better than in last years competition. With Brainstormers vs. WrightEagle we saw the same finalists for three worldcups in a row. The fact, that real innovations become sparse, indicates to me, that the problem of intelligent soccer is close of being solved for the very limited scenario found in the 2D Robocup simulation league. Maybe it’s time to focus on the more complex 3D scenario!?

For details on the final rounds of RC2008 see the DAINAMITE blog.

3 wins, 1 draw and 2 lost games. The DAINAMITE team arrives 3rd in the first Robocup 2008 round and thus reaches the quarter finals for the first time after two 9th places in the previous world cups. The next games start this saturday (chinese time). See this blog for more details…

We recently analized the structure and dynamics of social bookmarking systems. For this purpose, we spent six months on crawling about 142 million del.icio.us bookmarks coming from around 1 million users. I will give an overview of our findings here. Details can be found in this paper which I am going to present at the ECAI 2008.

The growth of del.icio.us

According to this blog, the del.icio.us bookmarking site went online in Sep. 2003 and, as our data indicates, has seen an exponential growth since then. According to our dataset, there where over 7,305,559 newly added bookmarks and 47,429 newly appearing del.icio.us users in December 2007.

There is an interesting period in the first half of 2006 where del.icio.us didn’t encounter much growth at all, but plateaud at around 3.5 million new bookmarks a month. This pattern was also reported by other authors, but the reason remains unclear.

The monthly growth of del.icio.us between 2004 and 2008 by posted bookmarks, new users, new URLs and new tags.

Read the rest of this entry »

Exactly one year ago, I was allowed to go to the Georgia Institute of Technology where the Robocup 2007 was held. Robocup is an AI testbed where robot teams of different types compete with each other in solving the problem of playing soccer. DAINAMITE, our 2D Simulation team, consisted of 11 java processes constantly sending “kick”,”turn” or “dash” commands to a simulation engine. These simple commands can result in a very realistic tactical behavior as can be seen from these logfiles (based on flash, a Second-Life visualization is under development).

In Atlanta we arrived 9th, missing the final rounds by one place and thus achieving exactly the same result as in the 2006 worldcup. I stopped my commitment to DAINAMITE in 2007 in order to focus on my PhD work. But, thanks to the work of the remaining and new team members DAINAMITE has continued to evolve. Robocup 2008 will be held in Sozhou, China, from July 14-20.

GOOD LUCK, DAINAMITE!!

Links:

DAINAMITE homepage (The team code is partly available for download (Java).)

Robocup2008 homepage

We decided to make all sources of the SPREE prototype described in the last blog publicly available under GNU AGPL. The project homepage providing the sources and documentation can be found here. The prototype was developed for and in collaboration with the Deutsche Telekom Laboratories using Python and Turbogears for rapid prototyping. The Server is powered by CherryPy and we decided on Genshi for dynamic templating and the powerful SQLAlchemy for ORM. SPREE is strongly AJAX, thanks to a combination of Dojo and MochiKit. As part of the code, we also release the SPREE iGoogle and Google Desktop widgets that inform SPREE users about incoming questions.

The sources allow you to create your own knowledge sharing community, including automatic question classification and expert finding, real time communication using web chat and rating functionality.

Try SPREE and let us know what you think of it.

During my PhD work here at DAI-Labor, we developed a prototype of a social search engine called SPREE. SPREE is intended as an online community where every user shares his knowledge and benefits from the knowledge of other users alike. To this purpose, the system automatically channels user questions to the most qualified experts. Techniques involved during the expert identification process include natural language processing as well as a tfidf based hierarchical classification of questions to a given knowledge taxonomy. The algorithms are described in our SPREE paper.

The prototype was written in Python using the Turbgears framework on the server side and a combination of Dojo/Mochikit for most javascript (AJAX) related stuff. It was much fun using Python/Turbogears for rapid prototyping after years of Java development.

You can register to SPREE at http://www.askspree.de. Let me know how you like it.