Jeffrey D. Kirtner, Steven L. Lytinen
This paper describes ULINK, a program designed to understand ungrammatical input. While most previous work in the field has relied on syntactic techniques or sublanguage analysis to parse grammatical errors, ULINK uses a semantics-driven algorithm to process such input. The paper gives a brief overview of LINK, the unification-based system upon which ULINK is built; special attention is given to those aspects of link which allow ULINK to use semantics to process ill-formed input. The details of ULINK’s algorithm are then discussed by considering two examples. The paper concludes with a discussion of related research and problems which remain to be solved.