TREC 2007 ciQA Track at RMIT and CSIRO

The only difference between all our system runs lies in the way in which the query was constructed. In our baseline system run, rmitrun1, we used those words within the brackets embedded in a question template as a query. In our second system run, rmitrun2, we added some additional words into the query; namely, those words form the narrative field of a question topic, except the introduction part (such as "The analyst would like to know of"). These words give elaborated information about what is counted as an answer. For the purpose of exploring the second research question, we would like to get an answer list from rmitrun2 that is sufficiently different from the one from rmitrun1, but not too dramatically dissimilar. We therefore experimented with giving different weights to the two sets of words from the title and narrative fields. Our submitted run used equal weight for the words from two fields.