Tuesday, July 21

SIGIR 2009 Day 1 Summary

Daniel has his summary of the day as well.

Monday Morning
I missed most of the morning talks yesterday, it was a frenetic rush as people were registering. I organized the layout of the poster session, which was non-trivial for the 100+ posters.

I heard good things about Paul's paper, Refined Experts: Improving Classification in Large Taxonomies.

I also want to check out Segment-Level Display Time as Implicit Feedback: A Comparison to Eye Tracking. I missed the session, but there's some coverage from CSAIL.

In the afternoon I attended the Retrieval Models I track. There were multiple papers that tried to address the risk/reward of ranking documents that are similar to one another. In short, if the most relevant (highest prob) pages are very similar, but are not the desired context that you can achieve better performance by introducing diversity into the results. I wonder if some simple document similarity/clustering models would be simpler and be just as effective.

One notable talk I heard about and want to read later is:
Enhancing Cluster Labeling Using Wikipedia by David Carmel

The evening was the poster session. There were an overwhelming number of posters and I know that I didn't get a chance to look at all of them in the depth that I wanted.

Marc, Elif, and I presented a poster Characterizing the Subjectivity of Topics. The goal is to better inform users about the broader context of the topic a given a single document (blog post) on the topic.

As usual, the best part was talking to people about their work.
Justin Zobel had a provocative poster Has Adhoc Retrieval Improved Since 1994? which they have in their evaluation workshop.

Another highlight was Omar's work using Mechanical Turk work.

The Glasgow guys had about half a dozen posters!

After the poster session Michael and a group of the Glasgow guys ended up at Uno's for drinks and ended up discussing our experiences with Hadoop and distributed computation.

No comments:

Post a Comment