Automated Evaluation Of Texts By Individual Teacher'S Model

We herein present a method by which to construct an automatic scoring system for compositions. The correctness ratio of the multiple-choice question is the most commonly used method for measuring a learners' progress in the current eLearning system. However, with respect to the eLearning system for composition education, it is not sufficient to use only the multiple-choice question to measure the ability of logical thinking and the power of expression. Descriptive-style questions are required in eLearning. Although a human teacher can grade tests and educate students, quickly grading and returning assignments and tests with appropriate explanations on a continuous basis is difficult for a large number of users. As such, in order to quickly grade and return assignments and tests and for fairness in evaluation based on stable criteria, an automatic scoring system is desired for grading descriptive-style questions in eLearning systems. In the English-speaking world, a number of studies have examined automatic text evaluation. However, metrics for essay evaluation are not opened in these systems or, when opened, language constituents to which the metrics relate are so abstract that learners are unable to improve their writing skills. We believe that systems with the above-described policy are inappropriate as a basis of eLearning for development of the learners' ability because such systems cannot separately describe the evaluation criteria to their learners. We herein present a model of evaluation that represents a human educator and a foundation for composition/short essay evaluation in an eLearning system, which describes the evaluation criteria for all language constituents. We also propose a method by which to manifest the evaluator model as weighting for metrics. The proposed methods make it possible to automatically score texts from broad viewpoints, to reveal linguistic factors used by individual evaluators, and to quantify the weights of the elements that contribute to the final score.