There is a nice article in the most recent issue of D-Lib Magazine called Enhancing Search and Browse Using Automated Clustering of Subject Metadata. The authors looked at ways to integrated automatic classification with traditional categories. “Results indicated that while the algorithm was somewhat time-intensive to run and using a local classification scheme had its drawbacks, precise clustering of records was achieved and the prototype interface proved that faceted classification could be powerful in helping end-users find resources.”
I like the practicality of this study. Lots of screens are shown, helping you grab onto issues discussed. They also talk about user testing and integrating that feedback into the designs.