2017 年 17 巻 p. 23-31
The authors have developed REX, a web application for facilitating English reading study, for non-native speakers of English. REX has a function that personalizes their individual readability equations based on their study logs and that predicts the ease or difficulty of the texts provided. For a parameter of such equations, the rate of difficult words, which does not appear in the specified vocabulary list, is often used, as in some of the well-known readability formulas. On the other hand, we believe that the use of a static vocabulary list to produce its rate does not contribute to personalization, because each learner has different vocabularies. This triggered the authors to incorporate a new function to let learners click their unknown words while they are reading texts, and further predict the status of known/unknown of the unclicked words. In this study, the practicality of the rates of difficult words is examined, which is computed from subjects' unknown word-lists in experiments.