Mathematics
and Computer Science Speaker Series
California State University, Stanislaus
Date: Tuesday, April 21,
2009
Time: 4:00 - 5:00 p.m
Room: P-101
Speaker: Melanie Martin
Title: Is what you're reading true?
Abstract:
This question, for print media, has been pondered for centuries
by scholars, philosophers, mathematicians, librarians and scientists,
among others. While the question may not actually be possible to
answer, some progress has been made. There is general agreement that
document attributes such as authority, coverage, objectivity, accuracy,
currency and support are useful in estimating the probability that
information in a document is true.
With the advent of the Web where virtually anyone can publish anything
anytime, we have vastly more information to sort through in our search
for truth. On one hand, it is not immediately obvious that the
attributes generally used for print media will serve equally well on
the Web. On the other hand, the Web can be viewed as a graph and the
topology of its links mined for new attributes.
In this talk, I will describe ongoing research on the creation of an
automatic measure of reliability (trustworthiness of information) in
the medical domain of the Web. I use a variety of features from
computational linguistics (syntactic and semantic), information
retrieval and web topology as input into machine learning algorithms.