Text classification for subjective scoring using K-nearest neighbors

Online testing is the main part in online learning, e-Learning. The modern problem is the answer scoring of assessments, tests, and quizzes. Subjective test is the complex question, which require human judgement evaluation. It requires extensive time to score the answers. Besides, in the large class, high number of learners, an instructor may unfair and difficult to score the subjective answer. In response, this paper presents text classification for subjective scoring using k-nearest neighbors in order to automatically score the subjective answer. The proposed approach makes use data mining techniques of k-nearest neighbors (KNN) algorithm to predict the score. The proposed algorithm splits text of subjective answer to many words/phrases using dictionary, which can cope with Thai and English languages. After that, KNN algorithm is applied with the proposed similarity algorithm. The proposed similarity is based on word matching and word ordering. If the pair of text match words/phrases and have the same order, the similarity is high. The results suggest the new approach is able to predict the subjective score and achieve the highest predictive accuracy than the standard classifiers. As a result, the proposed method can be a useful tool for instructors.