Generating Answers from a Linguistically Coded Knowledge Base

Abstract

The computational processing of natural language data is sometimes approached as a purely practical problem, for which ‘anything goes’ as long as the results are reasonably acceptable. Given sufficient ingenuity of the analyst, this approach may yield short-term successes, which will, however, be of limited theoretical interest. As a theoretical linguist, I am more interested in attempts at finding more principled, linguistically and psychologically motivated solutions to the problems involved. I also believe that such more principled solutions will yield better long-term practical results in any attempt to attack anything more complicated than the simplest natural language processing puzzles.