The Classification of Spoiler Comments in Thai Discussion Forums
Rangsipan Marukatat · 2020
Classifying comments about movies into spoiler or non-spoiler is a challenging application of text mining. This research proposes using intrinsic and extrinsic attributes for the classification task. Intrinsic attributes consist of words in a bag-of-words model and spoiler cues extracted from the comments, whereas extrinsic ones are genre information gathered from external sources. First, a few methods to select predictive words from the bag of words were compared. Ensemble attribute selector was found to achieve the best results. Then, the classification was done by support vector machine (SVM) and Naïve Bayes. The accuracies on unseen data were around 85--89% when using only bags of selected words, and close to 91% when using all the proposed attributes.