Text categorization for assessing multiple documents integration, or John Henry visits a data mine

Abstract

A critical need for students in the digital age is to learn how to gather, analyze, evaluate, and synthesize complex and sometimes contradictory information across multiple sources and contexts. Yet reading is most often taught with single sources. In this paper, we explore techniques for analyzing student essays to give feedback to teachers on how well their students deal with multiple texts. We compare the performance of a simple regular expression matcher to Latent Semantic Analysis and to Support Vector Machines, a machine learning approach. © 2011 Springer-Verlag Berlin Heidelberg.