Development of an Unsupervised LearningBased Automated Evaluation System of Descriptive Assessment

Abstract

Current research on automatic scoring using traditional supervised learning methods cannot grade responses to questions that are newly generated or created spontaneously. Additionally, relying on pre-developed questions and their corresponding scoring models for lesson planning may limit the creativity and diversity of instruction. This study proposes a method that can quickly evaluate student responses and generate feedback without the need for pre-developed models. We introduces the SAAI system, which employs unsupervised learning techniques to instantly create scoring models based on student responses, thereby generating evaluation and feedback information. The SAAI system complements the automatic scoring of traditional supervised learning methods and supports scoring for a wide range of newly generated questions. This research elucidates the principles and significance of this system.