Adult Content Classification on Indonesian Tweets using LSTM Neural Network

Abstract

In the last decade, social media networking sites have become an inseparable part of people’s life. However, not all of content in social media contain beneficial and necessary information. This can be seen from the existing of negative and harmful content in social media, such as adult or pornographic content. Therefore, this study aims to build a model for adult content classification by using Long Short Term Memory (LSTM) Neural Network to classify adult content and non-adult content. We also compared our LSTM methods with Multinomial Naive Bayes, Logistic Regression, and Support Vector Classification. According to our experiments, the best model was obtained from the LSTM model with two LSTM layers and dropout reached the accuracy of 98,39% and the loss value of 5,08&.