Knowledge Component Suggestion for Untagged Content in an Intelligent Tutoring System
Mario Karlovčec, Mariheida Córdova-Sánchez, Zachary A. Pardos
Tagging educational content with knowledge components (KC) is key to providing useable reports to teachers and for use by assessment algorithms to determine knowledge component mastery. With many systems using fine-grained KC models that range from dozens to hundreds of KCs, the task of tagging new content with KCs can be a laborious and time consuming one. This can often result in content being left untagged. This paper describes a system to assist content developers with the task of assigning KCs by suggesting knowledge components for their content based on the text and its similarity to other expert-labeled content already on the system. Two approaches are explored for the suggestion engine. The first is based on support vector machines text classifier. The second utilizes K-nearest neighbor algorithms employed in the Lemur search engine. Experiments show that KCs suggestions were highly accurate.
The final publication is available at Springer via https://doi.org/10.1007/978-3-642-30950-2_25.