Relationship between Student Writing Complexity and Physics Learning in a Text-Based ITS

摘要

In this paper we study 2217 essays written during ITS-based physics tutoring. Using output from the Stanford parser, we calculate various simple and more complex linguistic features, including average sentence length, tree height and number of subordinate clauses. Using the WEKA J48 implementation of the C4.5 algorithm and other statistics, we attempt to find relationships between linguistic features, the complexity of the students' text, students' scores on a physics posttest and their learning gain from the tutoring sessions.