engleski

Automatic extraction of learning concepts from exam query repositories

One of the biggest challenges in the process of establishing modern e-learning systems is figuring out ways to leverage legacy course materials and integrating them in the new information systems. Existing exam query repositories in particular are a very valuable data source, but one which usually lacks enough metadata to help establish relationships between exam questions and corresponding learning concepts whose adoption is being evaluated. In this paper we present the continuation of our research regarding the usage of educational data mining methods able to automatically annotate pre-existing exam queries with information about learning concepts they relate to. In our novel approach we leverage both textual and visual information contained in the queries. By combining the power of natural language processing which focuses on the text of the question, and annotated data extracted from figures accompanying the questions, we are able to further refine our classification methods and achieve noticeably improved results. By identifying learning concepts more accurately we further facilitate automatic creation of exams as well as even better insight into learning concept adoption. Our approach is again applied on data gathered from a large scale university course, and the results were validated in consultation with educational domain experts.

educational data mining ; exam queries ; learning concepts ; classification ; e-learning

nije evidentirano