Resolving Fragments in a Cardiovascular ITS

摘要

This paper describes an experimental approach to understanding the student input to our intelligent tutoring system, CIRCSIM-Tutor. The current parser, a cascade of finite state machines, handles the usual one and two-word answers that the students type in response to system questions extremely well, but there are four new kinds of input that we would like to be able to handle: answers to open questions (which are typically longer and more complex), student initiatives, hedges, and expressions of affect. The approach described here makes use of Copestake's Minimally Recursive Semantics in the context of Head-Driven Phrase Structure Grammar to analyze fragments and translate them into logic forms.