An automatic text comprehension classifier based on mental models and latent semantic features

Reading comprehension is one of the main concerns for educational institutions, as it forges the students' ability to comprehend and learn accurately a given information source (e.g. textbooks, articles, papers, etc.). However, there are few approaches that integrates digital sources of educational information with automated systems to detect whether an individual has comprehended a given reading task. This work main contribution is a text comprehension classification methodology for the detection of reading comprehension failures in educational institutions. The proposed approach relates situational model theories and latent semantic analysis from fields of psycholinguistics and natural language processing respectively. A numerical characterization of students' documents using structural information, such as the usage of text connectors, and latent semantic features are used as input for traditional classification algorithms. Therefore, an automated classifier is built to determine whether a given student could or not comprehend the information in the given stimulus documents. For the evaluation of the proposed methodology, using a set of stimulus documents, a set of questions must be answered by an experimental group of students. We have performed experiments using first year students from Engineering and Linguistics undergraduate schools at the University of Chile with promising results.