Mining Reading Comprehension Within Educational Objective Frameworks

In this paper we explore patterns in student behavior as they answer questions about documents they are reading. In earlier work [4] we showed that as students answer a question online, they can be categorized into one of 4 different clusters of “reading-scanning-scrolling” behaviors. Further, their readingscanning-scrolling behavior category predicts the quality of their answer to that particular question based on the level of that question in Bloom’s Taxonomy. We have performed a second experiment that confirms these earlier results. In a third exploratory experiment we also show how the reading-scanningscrolling clusters already discovered can be refined for use with another taxonomy, the Marzano Taxonomy. We are currently exploring whether other clusters can be found to help understand student behavior in terms of the Marzano Taxonomy.