Modeling and assessing student activities in on-line discussions

As web-enhanced courses become more successful, they put considerable burdens on instructors and teaching assistants. We present our work on developing software tools to support instructors by automatic assessment of pedagogical discussions. We are developing prototype measures of discussion quality that rely on the quantity of discussion contributions. We are also developing techniques for assessing discussion contributions automatically by mining discussion text. Using information retrieval and natural language processing techniques, our tools learn to detect the conversation focus of threaded discussions, classify topics of discussions, and estimate technical depth of contributions. The results from these assessment tools provide basis for the development of scaffolding and question answering techniques for pedagogical discourse.