The University of Massachusetts Amherst’s Participation in the INEX 2011 Prove It Track

We describe the process that led to the our participation in the INEX 2011 Prove It task. We submitted the results of six book page retrieval systems over a collection of 50,000 books. Two of our runs use the sequential dependency model (a model that uses both unigrams and bigrams from a query) and the other four interpolate between language model scores at the passage level and sequential dependency model scores at the page level. In this report, we describe our observations of these and several other retrieval models applied to the Prove It task.