In theory, I would have had more blog entries while at SIGIR last week. However, a number of things prevented me from doing anything:

1. Wireless barely worked during the conference;
2. The talks were interesting and I wanted to pay attention;
3. I decided to talk to other delegates during the evening;
4. I was lazy.

Luckily, there’s nothing like an 8 hour flight without network access to get things done.

Overall
SIGIR is clearly on the ascent. It’s not a huge shock… with all the competition in search and clear monetization, a ton of effort is going into the area from both academic and commercial interests. SIGIR now runs 3 paper tracks per session, and the proceedings clearly show that it’s now a much bigger conference than it was even 2 years ago. Last year’s SIGIR in Seattle drew record attendance, but people thought perhaps that was a fluke given how many Microsoft employees attended (over 100), given it was in Microsoft’s backyard. Well, this year was no different, and Amsterdam is quite a ways away from Redmond. Nonetheless, Microsoft still had about 50 employees present. Google had about 25, Yahoo maybe 20… so, clearly there’s some commercial interest.

Big Themes
There were a number of big themes at the conference this year. Gaining a fair amount of prominence were both Web IR (two sessions) and Learning to Rank (two sessions). There was also a lot more on Evaluation (three sessions) this year than in the past.

Papers
Here are a number of the papers I found interesting and would recommend reading. This probably shows more my biases in areas vs some ordering of SIGIR papers. Just as it was easiest, I’ve ordered these in the order they are in the SIGIR Proceedings.

Reliable Information Retrieval Evaluation with Incomplete and Biased Judgements [p. 63]
S. Büttcher, C. L. A. Clarke, P. C. K. Yeung (University of Waterloo)
I. Soboroff (National Institute of Standards and Technology)

Alternatives to Bpref [p. 71]
T. Sakai (NewsWatch, Inc.)

Fast Generation of Result Snippets in Web Search [p. 127]
A. Turpin, Y. Tsegay (RMIT University)
D. Hawking (CSIRO ICT Centre)
H. E. Williams (Microsoft Corporation)

The Influence of Caption Features on Clickthrough Patterns in Web Search [p. 135]
C. L. A. Clarke (University of Waterloo)
E. Agichtein (Emory University)
S. Dumais, R. W. White (Microsoft Research)

Information Re-Retrieval: Repeat Queries in Yahoo’s Logs [p. 151]
J. Teevan (Massachusetts Institute of Technology)
E. Adar (University of Washington)
R. Jones, M. A. S. Potts (Yahoo! Research)

Studying the Use of Popular Destinations to Enhance Web Search Interaction [p. 159]
R. W. White, M. Bilenko, S. Cucerzan (Microsoft Research)

Multiple-Signal Duplicate Detection for Search Evaluation [p. 223]
S. Huffman, A. Lehman, A. Stolboushkin, H. Wong-Toi, F. Yang, H. Roehrig (Google Inc.)

Robust Classification of Rare Queries Using Web Knowledge [p. 231]
A. Broder, M. Fontoura, E. Gabrilovich, A. Joshi, V. Josifovski, T. Zhang (Yahoo! Research)

Random Walks on the Click Graph [p. 239]
N. Craswell, M. Szummer (Microsoft Research Cambridge)

A Support Vector Method for Optimizing Average Precision [p. 271]
Y. Yue, T. Finley, F. Radlinski, T. Joachims (Cornell University)

A Regression Framework for Learning Ranking Functions Using Relative Relevance Judgments [p. 287]
Z. Zheng (Yahoo! Inc.)
H. Zha (Georgia Institute of Technology)
K. Chen, G. Sun (Yahoo! Inc.)

On the Robustness of Relevance Measures with Incomplete Judgments [p. 359]
T. Bompada, C.-C. Chang, J. Chen, R. Kumar, R. Shenoy (Yahoo!)

Test Theory for Assessing IR Test Collections [p. 367]
D. Bodoff (University of Haifa)
P. Li (State University of New York at Buffalo)

Strategic System Comparisons via Targeted Relevance Judgments [p. 375]
A. Moffat, W. Webber (The University of Melbourne)
J. Zobel (RMIT University)

Relaxed Online SVMs for Spam Filtering [p. 415]
D. Sculley, G. M. Wachman (Tufts University)

Know Your Neighbors: Web Spam Detection using the Web Topology [p. 423]
C. Castillo, D. Donato, A. Gionis, V. Murdock (Yahoo! Research Barcelona)
F. Silvestri (ISTI-CNR)

DiffusionRank: A Possible Penicillin for Web Spamming [p. 431]
H. Yang, I. King, M. R. Lyu (The Chinese University of Hong Kong)

HITS on the Web: How Does it Compare? [p. 471]
M. Najork (Microsoft Research)
H. Zaragoza (Yahoo! Research Barcelona)
M. Taylor (Microsoft Research)

HITS Hits TREC - Exploring IR Evaluation Results with Network Analysis [p. 479]
S. Mizzaro (University of Udine)
S. Robertson (Microsoft Research)

How Well Does Result Relevance Predict Session Satisfaction? [p. 567]
S. B. Huffman, M. Hochster (Google, Inc.)

Final Thoughts
The research going into SIGIR is really, really good and useful. This isn’t to say that previous work wasn’t; rather, SIGIR is moving into a world no longer completely dominated by the TREC competitive evaluation model and into a world where research is being done on real systems, collections, queries, and people. This is fantastic, as it will directly lead to improvements on systems that we all use.

This entry was posted 1 year, 2 months ago on Tuesday, August 7th, 2007 at 2:20 pm and is filed under Search and Industry. You can follow any responses to this entry through the RSS 2.0 feed.You can leave a response, or trackback from your own site.