Argument Identification in Indonesian Tweets on the Issue of Moving the Indonesian Capital
Amalia Huwaidah, Adiwijaya Adiwijaya, Said Al Faraby · Procedia Computer Science · 2021
Last October 2019, Indonesian Twitter community is busy discussing the issue of moving the capital city, and people are very eager to share their opinion in various expressions. This form of expression was alleged as a form of society expressing their opinions and arguments. This research uses a dataset from online discussions about moving Indonesian capital on Twitter. The goal of this study aims to identify whether a tweet contains argument or not. In this experiment, we use Multi-Class Support Vector Machine (SVM), and Multinomial Naïve Bayes (MNB) as the classifier and TF-IDF as feature extraction. Variation of Twitter data characters that have a lot of noise will be a challenge in this study so that some preprocessing processes will be carried out to overcome this problem. This research will investigate several combinations of preprocessing to discover the best result. We classify each tweet information such as argument, non-argument, and unknown. The best results with an accuracy of 71.42% were obtained by performing SVM with only a unigram feature. This study shows that the stopwords feature has effectiveness depends on which feature combination is implemented in the model.