An Attempt to Combine Features in Classifying Argument Components in Persuasive Essays

So far, several approaches have been done in detecting and classifying argumentation in persuasive essays. In this paper, we proposed some new features on top of the state-of-the-art researches in argumentation mining. We grouped 68 features into 8 categories; they are structural, lexical, indicators, contextual, syntactic, prompt similarity, word embedding, and discourse features. Instead of handcrafted features, we utilized word embedding as the feature. At the end of this paper, we presented the comparison between each group of features to classify the argument components. 402 persuasive essays were utilized. We found that structural features were the most significant feature while discourse features were not. After combining all features, we obtained 79.96% as the accuracy; it was slightly outperforming the state-ofthe-art accuracy which was 77.3%.