BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Random Walks on the Click Graph - Martin Szummer\, Microsoft Resea
 rch Cambridge
DTSTART:20071012T110000Z
DTEND:20071012T120000Z
UID:TALK8700@talks.cam.ac.uk
CONTACT:Johanna Geiss
DESCRIPTION:Search engines can record which documents were clicked for whi
 ch query\, and use these query-document pairs as ‘soft’ relevance judg
 ments. However\, compared to the true judgments\, click logs give noisy an
 d sparse relevance information. We apply a Markov random walk model to a l
 arge click log\, producing a probabilistic ranking of documents for a give
 n query. A key advantage of the model is its ability to retrieve relevant 
 documents that have not yet been clicked for that query and rank those eff
 ectively. We conduct experiments on click logs from image search\, compari
 ng our (‘backward’) random walk model to a different (‘forward’) r
 andom walk\, varying parameters such as walk length and self-transition pr
 obability. The most effective combination is a long backward walk with hig
 h self-transition probability. This is joint with Nick Craswell.
LOCATION:SW01 Computer Laboratory
END:VEVENT
END:VCALENDAR