華語作文分級系統

Abstract

有感於世界對於華語文學習的需求與日俱增，但在華語學習環境中，卻沒有像英文托福考試使用的e-Rater這類的工具，可以幫助華語文教師或學生進行教學或學習，因此研製一個給華語文領域使用的作文分級系統，希望能對此有所助益。 本論文之研究使用Stanford parser作為文法剖析器，開發出數個文法相關特徵，並以貝氏機率為機器學習之模型，實作出華語作文分級系統。 本研究所開發出的系統可達到93%的正確性，對於華語作文以分數作為分級的方法，可達到不錯的效果，亦可在華語老師的教學上或華語測驗裡實際使用。

Due to the learning boom of CFL, the needs of learning equipment increased. However, there is no such tool like e-Rater for TOEFL in the CFL learning field for Chinese teaching instructors and students to use. In this study, we tent to build up an automated essay scoring system for CFL learners leading to a better CFL learning environment. The system we developed in the study used Stanford Parser as a grammar parser to analyze and parse sentences to design some grammar features that could fit the system. We used Bayesian theorem as a machine learning model. By integrating features to the model, we built up a Chinese essay scoring system for CFL. The system could reach to 93% on the adjacent accuracy in rating the scores of essays and could literally use for the practical needs in CFL teaching or test.