Abstract
This study increased the accuracy and validated use of the Automatic Scoring System for Japanese Writing by Students of Japanese as a Second Language (JSL). Analysis was conducted using 107 writing samples from placement tests taken by JSL students. Statistics were extracted from the writing samples for variables such as the number of characters and number of sentences. Multiple regression analysis was then conducted using the statistics as explaining variables and scores from grading by experienced scorers as criterion variables, and a predictive equation was calculated. Furthermore, correlation between the results of grading by experienced scorers and the prediction results was evaluated using descriptive writing assignments given as part of placement tests in the second and third terms of 2000 in order to validate the predictive equation. As a result, it was apparent that the results of score prediction using the predictive equation were almost equal to the scores obtained from grading by experienced scorers.
