One of the great things about blogging as social medium is that you quickly discover the most reputable people in your field. When I started blogging and participating in other blogs that dealt with topics related to information retrieval, I quickly discovered Geeking with Greg.
Greg Linden was the primary developer and designer of the Amazon.com recommendations engine, which is one of the most widely used–and perhaps the most famous–recommendation system on the web. He tried his hand at a news and blog aggregation, Findory, that was based on collaborative filtering but ultimately gave up on that to join Microsoft Live Labs.
Greg has been blogging since 2004. His interests combine strong research interests with practical grounding. For example, he recently wrote about a paper at CIKM 2008 (by the way, he’s been great at blogging about the conference) on learning to rank using click data. Given my day job, I am sometimes involved in “bake offs” between our engine and those of competitors, and I’d be delighted to see prospects run evaluation experiments based on this paper.
Greg’s posts range from the theoretical (“Does the entropy of search logs indicate that search should be easy?”) or the practical (“Advertising, search, and drive-by malware“). He talks about great work coming out of Microsoft’s labs, but gives equal time to works from competing labs, such as Yahoo and Google.
In short, Geeking with Greg is a must-read for anyone serious about real-world information retrieval.