Computer-aided intervention for reading comprehension disabilities

Abstract

Our research work focuses on grouping of students based on error patterns in assessment outcomes for effective teaching of reading comprehension in early elementary education. The work can facilitate placement of students with similar reading disabilities in the same intervention group to optimize corrective actions. We collected ELA (English Language Arts) assessment data from two different schools in NY, USA, involving 365 students in total. To protect individual privacy of the participants, no background information that can possibly lead to their identification is collected for the study. To analyze underlying factors affecting reading comprehension without students’ background information and to be able to evaluate the work, we transformed the problem to a K-nearest neighbor matching problem—an assessment should be matched to other assessments performed by the same student in the feature space. The framework allows exploration of various levels of reading skills as the features and a variety of matching mechanisms. In this paper, we present studies on low-level features using the computer-generated measures adopted by literacy experts for gauging the grade-level readability of a piece of writing, and high-level features using human-identified reading comprehension skills required for answering the assessment questions. For both studies, the matching criterion is the distance between two feature vectors. Overall, the low-level feature set performs better than the high-level set, and the difference is most noticeable for K between 15 and 30.