Looking at things differently: Exploring perspective recall for informal text retrieval

When retrieving informal text such as blogs, comments, contributions to discussion forums, users often want to uncover different perspectives on a given issue. To help uncover perspectives, we examine the use of query expansion against multiple external corpora. We consider two informal text retrieval tasks: blog post finding and blog finding. We operationalize the idea of uncovering multiple perspectives by query expansion against multiple corpora from different genres. We use two approaches to incorporate these perspectives: as a rank-based combination of runs and a mixture of query models.
The use of external sources does indeed generate different views on a topic as becomes clear from the unique relevant results identified by the expanded runs compared to the baseline run. Even after combining the expanded run with the original run, unique relevant documents are found by both of the perspectives. As to the combination methods, the mixture of query models outperforms the rank combination, and leads to significant improvements in MAP score over the baseline.