An evolutionary system for pedagogical lexical analvsis

A system for automatically generating question and answer sessions from text can he useful in pedagogical environments. We have developed a system that uses pattern analysis text decompose text into distinct morphological Levels: paragraphs, sentences and words. Word recognition is carried out throug a neutral based generalization engine that repeatedly processes unknown words through B series of rules until the word is recognized. Experiments on this engine show differences in the effectiveness of this approach for varying neural net structures.