本篇論文提出了一個針對英語介係詞錯誤的偵測與更正系統,目的在於幫助ESL(English as second language)環境的英語學習者了解英語介係詞的用法。 本系統擁有幾項特質。首先,由於應用貝氏理論,因此本系統在效能上能有不錯的表現。再者,本系統應用了英語文法上的概念:content word與function word概念,因此我們可以觀察每個單字所包含的語意多寡是否會影響介係詞的選擇。 本系統包含了三種類型的演算法,運用統計模組、貝氏理論的基本演算法,根據套用不同出現機率的加權演算法,應用英語文法中content word、function word概念演算法。在數據討論我們將後二者演算法與基準演算法比較可發現在精確值與回饋值上的改進。
This paper presents a system within detection and correction for English preposition errors, the objective is to help the ESL (English as second language) learners in learning the usage of preposition. The system has some special features. First, it applies the Bayesian theory so it presents nice performance in efficiency. Second, it applies some concepts in English grammar: the content word and the function word, we can observe the influence on the choice of prepositions from each word’s semantic meaning. There are 3 types of algorithms in our system, the first is the baseline algorithm with statistical model using Bayesian theory, the second is the weighting algorithm depends on the kind of frequency it catches, the third is the algorithm applying the concept of content word and function word. In the Discussion we compare latter two algorithms with baseline algorithm and we can find some improvements from examining precision and recall.