Handling ill-formed natural language input for an intelligent tutoring system
Yoonhee Lee, Martha W. Evens · 1990
This research focuses on the design and development of a subsystem for handling ill-formed input in an intelligent tutoring system. For this purpose, I built the Ill-Formed Input Handling System (IFIHS). The IFIHS system is designed for understanding ill-formed natural language input from first year medical students. The IFIHS system consists of a screen manager, spelling corrector, parser, and understander. The IFIHS system is implemented on a Xerox 1108 AI Machine in the Interlisp-D language. Fast response time and user friendliness were the most important considerations in the design. The screen manager provides a full screen editor and input tracer for an easy editing capability for the student. I made the screen manager as simple as possible because most medical students have little computer experience. My spelling correction program can handle most kinds of spelling errors including character case, order reversal, missing characters, added characters, and character substitutions. This spelling correction system also corrects unexpected abbreviations and word boundary errors without making extra lexical entries. This spelling correction system can detect most student spelling errors and correct them automatically. We stored the lexicon in a trie for fast searching. Each lexical entry contains orthographic, syntactic, and semantic information so that the system can retrieve information for both spelling correction and the parser/understander at the same time. The LFG parser and the understander are designed to handle most types of simple sentences, and some kinds of fragments, ellipses, anaphora, and semantic corrections. The LFG parser produces the c-structure and f-structure for each input sentence in turn. The understander produces the logical form using the f-structure and the dialog history, and passes the logical form to the controller of the intelligent tutoring system.