CyberiaThe University of Virginia Information Retrieval Group
A major part of our work is research in distributed Information
Retrieval. We are developing new techniques aimed at constructing
a user-customized Personalized Information Environment
(PIE) that can be
tailored to the needs of specific users.
We have recently constructed a testbed for evaluating distributed information retrieval approaches and are currently working to compare different approaches using this testbed.
When dealing with distributed collections, different collections may employ different conventions for naming entities in the collection. This can adversely affect retrieval performance by obscuring similarities among those entities. This problem can be lessened by the use of authority files which list accepted variants of names; however, the construction of these authority files is labor-intensive. We have been studying the potential for partially automating this process, as in the EVOC Project.
We are also investigating the use of multiple
viewpoints of a document collection during searching or browsing
to facilitate the location of useful documents. Our research in
this area is focused on specific techniques for interacting with
document collections. We are investigating the representations,
framework and processes required to support these interactions.