Six good predictors of autistic reading comprehension

Recent Advances in Natural Language Processing (RANLP 2015)This paper presents our investigation of the ability of 33 readability indices to account for the reading comprehension difficulty posed by texts for people with autism. The evaluation by autistic readers of 16 text passages is described, a process which led to the production of the first text collection for which readability has been evaluated by people with autism. We present the findings of a study to determine which of the 33 indices can successfully discriminate between the difficulty levels of the text passages, as determined by our reading experiment involving autistic participants. The discriminatory power of the indices is further assessed through their application to the FIRST corpus which consists of 25 texts presented in their original form and in a manually simplified form (50 texts in total), produced specifically for readers with autism.University of Wolverhampto