Automated Essay Scoring Using the KNN Algorithm

The K-Nearest Neighbor (KNN) algorithm for text categorization is applied to CET4 essays. In this paper, each essay is represented by the vector space model (VSM). After removing the stop words, we chose the words, phrases and arguments as features of the essays, and the value of each vector is expressed by the term frequency and inversed document frequency (TF-IDF) weight. The TF and information fain (IG) methods are used to select features by predetermined thresholds. We calculated the similarity of essays with cosine in the KNN algorithm. Experiments on CET4 essays in the Chinese Learner English Corpus (CLEC) show accuracy above 76% is achieved.