TICL is a simple stripped down version of something we have been hacking recently. There has long been the need for a text classification workbench felt by researchers and practitioners. I proposed some of these things in my thesis looking forward chapter. Some recent research (including ours), industrial work, and experience has also indicated the usefulness of something along these lines. Now that I am not a student anymore, I could talk of business value and applications but I won't ;-). There is good value in real-world settings and I leave it at that.
While we have a more integrated and richer system going internally, a bare-bones first step is available as TICL. It works on Tomcat with a few downloadable additional jar files, works off Rainbow's directory structure, and has a simple naive Bayes classifier implemented. Database support, support for other input/output formats may follow. Hopefully someone somewhere will find it useful. All comments are welcome.