UWaterlooMDS at the TREC 2019 Decision Track

Mustafa Abualsaud, Fuat Can Beylunioglu, Mark D. Smucker

Abstract:

In this report, we discuss the experiments we conducted for the TREC 2019 Decision Track. This year, our goal was to
investigate the effect of document credibility on the quality of automatic runs. To address credibility, we combined scores
from a spam classifier and a credibility classifier trained to detect non-trustworthy websites. The results from both classifiers
were then used to modify a baseline BM25 ranking. In addition to the automatic runs, we also submitted manual runs using
the HiCAL system. Our manual runs modify a baseline BM25 ranking using manually judged documents found using the
system.