Nagaoka University of Technology -> Muneo Kitajima -> Publications -> Here

Last updated: 24-April-2026

--

Nakahira, K.T. and Kitajima, M. (2026)

Proposal of a Semi-Automatic Classification Method for Estimating Conceptual Understanding in Short Answer Grading for Semi-Open-Ended Questions Using Word Co-occurrence Networks

One of the effective method for estimating learners' level of understanding of acquired concepts involves using semi-open-ended questions. The method is well known to be particularly beneficial for questions requiring scientific explanations, and it is adopted in large-scale academic achievement tests and trend assessments. On the other hand, Short Answer Grading often relies on manual scoring by markers, raising concerns about workload and the depth of insight into specialized knowledge. To alleviate marker workload, automated evaluation combining natural language processing and machine learning, or utilizing generative AI, is anticipated. However, introducing machine learning requires large amounts of training data and also faces issues related to the native language of the test takers. As one solution to these problems, a method that enables classifying learners' understanding levels with minimal effort based on collected short answers, even under conditions of limited information for machine learning, is also considered beneficial. In this paper, we propose a method that classifies short answers into three levels based on conceptual understanding depth as one approach to short answers grading for semi-open-ended questions. The proposed method applies degree analysis, well-known in word co-occurrence networks.

Download

• pdf
• 2,800K

Nagaoka University of Technology -> Muneo Kitajima -> Publications -> Here

mkitajima@kjs.nagaokaut.ac.jp