AAAI Spring Symposium on
Machine Learning in Information Access
Suggestions for Research Projects

Important (or at least Fun) Problems
(From David Lewis, lewis@research.att.com)

1. Structure
   • within docs (passages, extracting facts,...)
   • between docs (redundancy, discourse,...)
2. Time
   • change in user interests / concept drift
   • change in distribution of docs processed
3. People
   • integration of user knowledge and learning
   • integration of community knowledge and learning
4. Trust
   • accurate estimation of class membership probability
   • better theoretical understanding of learning rate, etc.
5. Diversity
   • combining/managing multiple information sources
   • combining/managing multiple information systems