We describe the goals, architecture, and functioning of the TRAINS-93
system, with emphasis on the representational issues involved in putti
ng together a complex language processing and reasoning agent. The sys
tem is intended as an experimental prototype of an intelligent, conver
sationally proficient planning advisor in a dynamic domain of cargo tr
ains and factories. We explain some of the goals and particulars of th
e KRs used, evaluate the extent to which they served their purposes, a
nd point out some of the tensions between representations that needed
to be resolved. On the whole, we found that using very expressive repr
esentations minimized the tensions, since it is easier to extract what
one needs from an elaborate representation retaining all semantic nua
nces, than to make up for lost information.