Switching colours

As of this month, I am a postdoc at the Database Systems research group, headed by Prof. Kjetil Nørvåg at the Norwegian University of Science and Technology (NTNU) in Trondheim, Norway. I would like to say a big thank you to all my former colleagues in Amsterdam for providing an extremely friendly and inspiring research environment throughout the past several years. I wish you best of luck, and hope to see you at the next conference!

My research interests remain essentially unchanged: capturing, representing, and organizing information related to entities, in semantically meaningful ways. And, big data, of course.

Work hard, relax hard!

This picture gives you a bit of an idea how my week in the run-up for SIGIR deadline looked like (I also have a nice collection of energy drink cans at the office). But, that is finally over, and I managed to get in my submissions on time. That means I’m done with the work hard part, now it’s time for some serious relaxing and recharging of batteries.

I’ll be on vacation until Feb 10, and this time, I won’t be checking my emails. If anything urgent comes up… that has to wait, I’ll deal with it when I’m back.
I’m off to relax!

Looking back on 2009 and forward to 2010

This year has started quite intensely, with a research grant proposal deadline already on week 1. But, I’d like to take a moment and look back on 2009, before rushing on to the next deadline (which is SIGIR 2010, in less than a week away).

First, I’d like to make an honorable mention of four of my colleagues/co-authors who have defended their PhD and became doctors in 2009. They are (in temporal order):

They all did a great job, congrats!

A significant milestone of 2009 was the launch of the TREC Entity track. The overall aim of this new track is to perform entity-related search on Web data. The track defines entities as “typed search results” or “things”, represented by their homepages on the web. In other words, our working definition of an entity is “something with a homepage”, and searching for entities thus corresponds to ranking these homepages. (As a sidenote: I am well aware that this definition of an entitiy is far from perfect, yet, the URL of a homepage is the best entity identifier we could come up with so far.)
The first year of the track investigated the problem of related entity finding:

Given an input entity, by its name and homepage, the type of the target entity, as well as the nature of their relation, described in free text, find related entities that are of target type, standing in the required relation to the input entity.

This task can be seen as a mixture of Question Answering (specifically, the QA list task) and homepage finding. In the first year, we limited the track’s scope to searches for instances of the organizations, people, and product entity types.
Thirteen groups participated and submitted a total of 41 runs; this demonstrates clear interest, and, I think, is quite decent for the first edition of the track (beating well established tracks, like the Blog track, in terms of the number of participating teams).
The track continues in 2010, where the main task will (again) be related entity finding, with moderate changes and more topics. We are also planning to feature another subtask (more details will follow on the track’s mailing list).

Another important development was the release of the EARS toolkit. EARS stands for Entity and Association Retrieval System, and is an open source implementation of entity-topic association finding models, used so far mostly in the context of expertise retrieval, but also for other tasks, for example blog distillation. While currently the functionality of EARS is limited to two baseline models (“Model 1″ and “Model 2″), a number of additions are coming in future releases, throughout 2010. Most notably, proximity-based variations of existing models, and methods for finding entity-entity relations (i.e., addressing the related entity finding task defined above).

A yearly evaluation would not be complete without mentioning citation counts. Some say, citation is to publication as price is to stock. So, I do it this time in a NASDAQ-stlye. Total citation count has doubled (from 214 to 433) and my H-index has also increased, from 8 to 11. The top performing papers are shown below.

Citation counts (as of Jan 1) 2009 2010 +/-
1. Formal models for expert finding in enterprise corpora (SIGIR 2006) 74 136
2. Finding experts and their details in e-mail corpora (WWW 2006) 27 40
3. Broad expertise retrieval in sparse data environments (SIGIR 2007) 13 36 Up
4. Determining Expert Profiles (With an Application to Expert Finding) (IJCAI 2007) 10 25 Up
5. Why are they excited? identifying and explaining spikes in blog mood level (EACL 2006) 13 21
6. Language Modeling Approaches for Enterprise Tasks (TREC 2005) 16 20 Down

+/- denotes the position change in the relative ordering of my papers, according to the number of citations.

According to these numbers, our SIGIR 2006 paper is (still) a massive leader, and keeps on following the rich-getting-richer trend. A more interesting observation is that our expert profiling work seems to have gained impact and attention in the past year, as citation counts for the two profiling papers have almost tripled in 2009. This is good news, especially in the light of some ongoing work we are performing in this area.

That’s it for now (longest post ever). I wish a successful 2010 to everybody (and good luck to those with a SIGIR deadline)!

Last bundle of updates for 2009

I haven’t had time to post entries on my blog over the past few weeks (or even months — has it really been that long ago?). Anyway, here is a couple of things worth mentioning before 2009 is officially over.

A newer version of the EARS toolkit has been released. Major changes concern document-entity associations and faster computation of candidate models, as well as support for MS Visual Studio. See the changelog for details.

Our paper entitled Category-based Query Modeling for Entity Search, with Krisztian Balog, Marc Bron, and Maarten de Rijke as authors, has been accepted to ECIR 2010 and is available online now.

Abstract. Users often search for entities instead of documents and in this setting are willing to provide extra input, in addition to a query, such as category information and example entities. We propose a general probabilistic framework for entity search to evaluate and provide insight in the many ways of using these types of input for query modeling. We focus on the use of category information and show the advantage of a category-based representation over a term-based representation, and also demonstrate the effectiveness of category-based expansion using example entities. Our best performing model shows very competitive performance on the INEX-XER entity ranking and list completion tasks.

See also: ECIR 2010 accepted papers, posters, and demos.

The TREC Enterprise 2008 overview paper has finally been posted to the proceedings.

Happy 2010!