Machine Learning Questions in Information Retrieval
Prof. Richard K. Belew
University of California at San Diego
2:30pm Friday April 28 in Room 2310
Finding samples of free-text "about" topics of interest has been an
important problem for computer science for a long time, but with the
increasing size of our hard-disks, and especially as the world
discovers the Internet and imagines it all as one great Digital
Library, the critical importance of this issue is becoming apparent to
all. Beyond the importance of this problem domain, I will argue that
free-text information retrieval is especially amenable to machine
learning techniques. The key concept linking these two fields is
"relevance feedback," information collected from users as a
natural byproduct of their browsing behaviors. The talk will focus
on this source of information as a training signal for machine
learning algorithms. It is intended to be a general survey
of techniques used in our laboratory and elsewhere, as well as
outstanding research issues.