Measuring Hint Level in Open Cloze Questions

Providing the first few letters of a missing word in a sen- tence gives information about this word. This paper attempts to measure the information transmitted in that case. In or- der to do so, we analyzed response accuracy for open cloze questions, that is fill-in-the-blank questions without multiple choice answers. In this study, native and non-native speak- ers of English answered a series of open cloze questions that were semi-automatically generated. Hints were provided that consisted of the first few letters of the missing word. Results showed that question difficulty, hence the quantity of infor- mation transmitted, is related to the number of letters that are provided, to physical properties of these letters and to sylla- bles formed by these letters. Performances did not appear to depend on letter or syllable frequency. Controlling hint level in a word completion task is critical in order to provide prac- tice exercises adapted to student levels.