RecSys09

A couple of weeks ago I attended the 3rd ACM Conference on Recommender Systems which was an experience nothing else than excellent.

Day one started with a tutorial on trust based recommendations bu Jenifer Golbeck who is an authority when it comes to trust and recommendation systems. Luckily she was kind enough to put the contents of the tutorial online, available here.
In the tutorial we were presented with algorithms, techniques and other interesting (and nerdy) stuff. Since I'm currently working with (a kind of) trust in the field of movie recommendations this was extremely interesting. Too bad this wasn't an all day tutorial.

The second tutorial was on Bayesian modelling in recommender systems by Asim Ansari, unfortunately I ended up missing the first two-thirds of this tutorial so can't really give a recap on this one.

The third tutorial was on Important Research Challenges in Recommenders and was given as a panel discussion hosted by John Riedl and backed by Jon Sander from Netflix (the prizemaster himself) and Todd Beaupre from Yahoo!
The discussions in during this tutorial were all more or less leading to the final event of the day were the Netflix Prize winners presented their approach(es).
One of the memorable quotes during the netflix discussion was Jon Sanders starting his talk with "those of you who don't know what the Netflix prize is, you're at the wrong conference". Another memorable, and quite interesting, discussion was wether or not meta data can improve recommendations or not. This was an extra hot topic since one of the winner-teams had a paper titled "Recommending new movies: even a few ratings are more valuable than metadata" that they were going to present the following day at the conference. Anyhow, the full set of winners were convinced meta data was not that important, in the context of the Netflix prize. This was something many of the listeners had a hard time accepting..

Anyhow, I'm not going to give a full recap of the conference, as there have been numerous others who have done that, this one being one of the better ones.

There was a lot of twittering going on during the conference, and #recsys09 was a really hot topic during a couple of days. One thing this resulted in was the140 character keynote summary by Neal Lathia (also mentioned at the Strands blog)