When is a union really an intersection? Problems interpreting reference to locations in a dialogue system

This paper describes issues that arose in our implementation of an interpreter for locative expressions in a spoken dialogue system. The expressions involve complex adverbial modification, include imprecise and innacurate content, and reflect conventional practices specific to the way English speakers refer to roads and highways. Our system utilizes semantic features in the parser to perform disambiguation, and domain-specific reasoning to resolve variables in the logical form to the correct referent. The functionality implemented to date was evaluated against a small corpus of naturally-produced expressions, of which 46% were correctly interpreted.