Robust natural language parsing in computer-assisted language instruction

Abstract

We discuss the role of robustness in programs for computer-assisted language instruction which attempt to parse natural language input. A program has been implemented which checks for certain grammatical errors in French sentences entered by a student. Because errors are expected the system is built around a robust parser. We briefly outline the Language PROLOG in which the system is implemented and then describe the system, demonstrating the techniques which it uses for robust analysis.