Deep Learning-Based Implementation of Hate Speech Identification on Texts in Indonesian: Preliminary Study

This paper presents an implementation of hate speech identification task for text data written in Indonesian language. There are some studies purposed for similar problem, but all of them use classical machine learning approach, whose heavily depends on the feature engineering. Switching the domain of data set means that the feature engineering should be redone. To address this issue, this preliminary research proposes another method based on deep learning approach which needs no feature engineering and is also adaptive to the varying context. Using data sets sourced from Twitter posts, the proposed method gives better result of 94.5% F1-score at a minimum.