Automatic and Semi-Automatic Skill Coding With a View Towards Supporting On-Line Assessment

This paper explores the problem of automatic and semi-automatic coding of on-line test items with a skill coding that allows the assessment to occur at a level that is both indicative of overall test performance and useful for providing teachers with information about specific knowledge gaps that students are struggling with. In service of this goal, we evaluate a novel text classification approach for improving performance on skewed data sets that exploits the hierarchical nature of the coding scheme used. We also address methodological concerns related to semi-automatic coding.

• Type: Conference
• Year: 2005
• Where: AIED
• Authors: Carolyn Penstein Rosé, Pinar Donmez, Gahgene Gweon, Andrea Knight, Brian Junker, William W. Cohen, Kenneth R. Koedinger, Neil T. Heffernan
• Keywords: AIED 2005, Artificial Intelligence, On-line Test Items, Overall Test Performance, Semi-automatic Coding