Using Students' Performance to Improve Ontologies for Intelligent E-Learning System

Abstract

Ontologies have often been recommended for E-learning systems, but few efforts have successfully incorporated student data to represent knowledge conceptualizations. Defining key concepts and their relations between each other establishes the backbone of our E-learning system. The system guides an individual student through his/her course by evaluating their progress and suggesting instructional material to review based upon their answers. Three main tasks are performed within this framework: building ontologies for the course, measuring a student's understanding level for the concepts, and making personal suggestions to create an individualized learning environment. This paper presents: the integration of ontologies, assisted with student data, together with an intelligent Recommendation Module for the development of an E-learning system; the comparison and correction adaption of ontology from students' mind maps; and the assessment of students' actual weaknesses in comparison to what Recommendation Module suggests. The sample of 127 students, five classrooms, was conveniently selected among seventh grade students of a demographically average school in a major city in Turkey. The students' achievement was assessed and the scores for different questions were investigated for associations with concepts made in the students' minds. The results provided significant correlations among scores, and a fit model for the concepts represented by questions. The student suggested model slightly differed from the ontology map from the experts. Based on the data-supported model, the Recommendation Module more accurately determined the students' learning deficiencies and suggested concepts to be reviewed.