A Similarity-Calculation Method of Geometric Problems for Adaptive e-Learning

Abstract

In the recent years, increasing attention has been drawn toward adaptive e-learning, which is a learning method that uses educational big data to flexibly change the learning content according to the proficiency of a learner. In this study, we proposed an overview of an adaptive e-learning system that focuses on the solution procedure, which comprises the knowledges and operations used by a learner to solve a problem. Using the example of elementary geometric problems, we develop a prototype of our proposal in which the situations of learning are formalized on the basis of an adaptive e-learning context model, which is represented using a meta-network. This prototype comprises three subsystems. One is an expert system that automatically generates various solution procedures for a given problem. The other two are an inference system that identifies the procedures that a learner is using by calculating their similarity with the procedures generated by an expert system and a classification system of given problems by calculating the similarity of the procedures of each problem; the similarity is calculated using both the Levenshtein distance and Needleman–Wunsch algorithm, respectively. Compared with conventional methods, the proposed system might provide more detailed support for learners.