Reading errors made by skilled and unskilled readers: evaluating a system that generates reports for people with poor literacy

This study evaluates a natural language generation system that creates literacy assessment reports in order to create more readable documents. Prior research assessed comprehension and reading speed on modified documents. Here, we investigate whether individuals make less reading errors after the system modifies documents to make them more readable. Preliminary results show that poor readers make more errors than good readers. The full paper will describe readers' rates of errors on documents modified for readability