We did great in terms of having the most flexible system (any of E, F, G retrieve any of E, F, G). Only one other group did as well: ETH, the ones running the track.
Our performance was not stellar. We didn't use any linguistic information nor any modern IR ideas. We have lots of excuses, but we need to do more tests to see if we can improve things.
The results look reasonable (try ``oil crisis inflation''), but average precision was well below that of the other participants. A demo is available from the U. Colorado.
