| Summary: | 碩士 === 國立高雄應用科技大學 === 資訊工程系 === 99 === Automatic Chinese essay scoring (ACES) is a highly important research tool for educational tests and course teaching, and is used in various fields such as psychometrics. Traditional methods of manual scoring require a significant amount of time and high costs; therefore, numerous studies have proposed automatic essay scoring methods. Most of these methods use various writing characteristics, such as the number of words, number of topics discussed, and the degree of colloquialism in the essay, as the scoring criteria. A valid mathematical model then integrates the various characteristics demonstrated in the essay and calculates a possible score. Therefore, the writing characteristics selected as the scoring criteria, and the chosen mathematical model, have a significant impact on the resulting score. Though the traditional forecasting model integrates the preceding characteristics satisfactorily, it does not perform well with the writing characteristics discovered in recent years. For example, writing characteristics such as the use of rhetoric and the comparison of the overall essay structure cannot be assessed using a continuous numerical value; and the traditional forecasting model has difficulty integrating different types of numerical data. Therefore, this study proposes a new method that uses the existing fuzzy rules to automatically generate fuzzy inference rules to integrate writing characteristics with different numerical value types. The experiment results show that the scoring method proposed by this research integrates traditional characteristics and various numerical value types of characteristics in a way similar to the existing model. In addition, because the scoring process of the proposed method is more similar to manual scoring by human experts, the resulting scores are more easily accepted by the public. Therefore, for high-stakes tests, the proposed method has more viability than the traditional scoring model.
|
|---|
