Revealing essential notions: an algorithmic approach to distilling core concepts from student and teacher responses in computer science education

Abstract

Purpose: This study aims to assess subjective responses in computer science education to understand students' grasp of core concepts. Extracting key ideas from short answers remains challenging, necessitating an effective method to enhance learning outcomes. Design/methodology/approach: This study introduces KeydistilTF, a model to identify essential concepts from student and teacher responses. Using the University of North Texas dataset from Kaggle, consisting of 53 teachers and 1,705 student responses, the model's performance was evaluated using the F1 score for key concept detection. Findings: KeydistilTF outperformed baseline techniques with F1 scores improved by 8, 6 and 4% for student key concept detection and 10, 8 and 6% for teacher key concept detection. These results indicate the model's effectiveness in capturing crucial concepts and enhancing the understanding of key curriculum content. Originality/value: KeydistilTF shows promise in improving the assessment of subjective responses in education, offering insights that can inform teaching methods and learning strategies. Its superior performance over baseline methods underscores its potential as a valuable tool in educational settings.