An Open-Ended Question Self-Explanation Classification Methodology for a Virtual Laboratory Learning System

Scientific experiments are essential for science and technology education. Experiments in laboratory cost materials, require preparations, and sometimes cause hazards. A widely used educational tool with many advantages, e.g. cheap, repeatable, suspendable, and safe, virtual laboratory has gradually become a major experimental tool in most elementary and high schools. In educational science experiments, one major challenge is how to initiate students on scientific inquiry and ensure there are multiple opportunities for their formative self-assessment and revision. The self-explanation strategy has proven effective in deepen students' understanding of the concepts they are trying to learn. Using self-explanation strategy in educational science experiments might be an effective way to help students think about the observed results of science experiments and build correct scientific concepts. On the other hand, researches point out that using open-ended questions is better than traditional multiple-choice questions for self-explanation strategy. But when using open-ended question self-explanation strategy, without proper prior knowledge and guidance, a student may go wrong in the processes of deduction and result in constructing misconceptions that will become obstacles in further knowledge constructions. Therefore, a learning system that uses open-ended question self-explanation strategy should give proper feedback in order to help students build correct concepts when in self-learning mode. To help students operating in virtual science laboratory and constructing correct concepts from observed results this study constructs an online virtual laboratory learning system with open-ended question self-explanation strategy and proper feedback for natural science course of primary schools. The system uses natural language processing (NLP) technology to analyze students' self-explanation strings, compares the results with coded classification rules, established by an expert from reference explanations, to check the correctness of the strings and possible misconceptions in them, and gives proper learning material, as feedback, for the students to revise possible misconceptions. In the final experiment, the system records and checks all self-explanation strings from 53 students and gives them proper feedback, which reaches an average accuracy of 84.45% after the expert verify the results.