Simulating misreading

Abstract

Physical misreading (as opposed to interpretational misreading) is an unnoticed substitution in silent reading. Especially for legally important documents or instruction manuals, this can lead to serious consequences. We present a prototype of an automatic highlighter targeting words which can most easily be misread in a given text using a dynamic orthographic neighbour concept. We propose measures of fit of a misread token based on Natural Language Processing and detect a list of short most easily misread tokens in the English language. We design a highlighting scheme for avoidance of misreading.