University of Wolverhampton at the TREC 2011 Microblog Track.
George Paltoglou, Mike A Thelwall · Text REtrieval Conference · 2011
In this report we discuss the experiments we conducted at the University of Wolverhampton for the Microblog Track at TREC-2011. As this was the first time we participated in TREC and the particular task presents some unique challenges we initially focused on properly analyzing and indexing the new Tweets2011 Corpus. We experimented with the effects that some standard IR techniques, such as query expansion and proximity models have in this setting. Initial results indicated that both techniques provide small increases in precision, but more experiments are needed for final conclusions to be reached. Lastly, we experimented with using the page that a tweet links to as part of the tweet. The results were particularly low, indicating a potential error in the indexing process or a natural outcome, due to the increased length of the combined documents. More research into answering the issue is underway.