Automated Essay Scoring For Testing Chinese As A Second Language

With the development of computer science and technology, Computer Based Test and the Computer Adaptive Test have already come true. This resulted in a study on computerized scoring .Many researchers have been keeping on investigating automated essay scoring for decades. Most current approaches to the automated essay scoring are for English, however, none for Chinese .Even the research on this field is seen all over China.Purpose of this paper is to make an automated essay scoring system for testing Chinese as a second language .We took MHK (third level) essays as researching samples .Sample 1 included 583 essays on 7 different topics, and sample 2 included 488 essays on a same topic .We chose multiple linear regression as the research method . 45 measurable features that may reflect writing qualities are attracted from the essay texts. Then we took all the measurable features as independent variables, the essay scores given by human raters as dependent variables, and analyzed them under stepwise regression and enter regression (two different methods in withdrawing variables when doing multiple linear regression).Finally we chose from 8 regression equations ,and decided the regression equation "mhk₅=0.983+7.798E-02×mhk_shum+4.301E-03× zizong—2.182E-02 × cuoyufa+1.077E-02× ziyb" as the ideal tool to scoring essay automatically. At last, we used regression equation "mhk₅" to score all of the essays in sample 2, and achieved a result reaching a Pearson correlation as high as 0.566 with human raters.