Human words must be rapidly recognized by a computer system designed to maintain an interactive dialogue with humans. This paper discusses features of the PLATO IV-TUTOR system which facilitates the matching of words and concepts offered by students to those proposed by lesson authors. The first section examines a word recognition algorithm that handles spelling errors without doing character-string manipulations or linear searches. The spelling algorithm uses human rather than machine criteria. The second section describes a method that successfully obtains the human concept from a multiple word input (e.g., a sentence) without doing tedious syntactic parsing. (Author/TS)