Online Programming Semantic Error Feedback using Dynamic Template Matching

Abstract

Many of automated computer programming feedback is generated based on static template matching that need to be provided by the experts. This research is focusing on developing an automated online programming semantic error feedback by using dynamic template matching models based on students' correct answers submission. Currently, there is a lack of research using dynamic template matching model due to their complexity and varies in terms of programming structure. To solve the formulation of the dynamic templates, a new automated feedback model using front and rear n-gram sequence as the matching technique was developed to provide feedback to students based on the missing structure of the best-matched template. We have tested 60 student's Java programming answers on 3 different types of programming questions using all the dynamic templates randomly chosen for each student. An expert was assigned to manually match the student's answer with the 3 randomly chosen templates. The result shows that 80% of the best-matched templates for each student using the technique were similarly chosen by the expert. Based on the matched template, the student will be given feedback notifying the possible next programming instruction that can be included in the answer to get it correct as was achieved by the template. This model can contribute to automatically assist students in answering computational programming exercises.