Are automatically identified reading strategies reliable predictors of comprehension?

Abstract

In order to build coherent textual representations, readers use cognitive procedures and processes referred to as reading strategies; these specific procedures can be elicited through self-explanations in order to improve understanding. In addition, when faced with comprehension difficulties, learners can invoke regulation processes, also part of reading strategies, for facilitating the understanding of a text. Starting from these observations, several automated techniques have been developed in order to support learners in terms of efficiency and focus on the actual comprehension of the learning material. Our aim is to go one step further and determine how automatically identified reading strategies employed by pupils with age between 8 and 11 years can be related to their overall level of understanding. Multiple classifiers based on Support Vector Machines are built using the strategies' identification heuristics in order to create an integrated model capable of predicting the learner's comprehension level. © 2014 Springer International Publishing Switzerland.