Indonesian Twitter Data Pre-processing for the Emotion Recognition

Abstract

The 2019 Presidential Election in Indonesia causes sharp political polarization. The battle of discourse between two massive camps took place on social media. Various aspects of public opinion showing how people think and act can be found easily. Twitter as the most popular microblogging platform, offers a place to express a variety of thoughts and opinions. This makes Twitter as a source of opinion mining that can be used to detect people's emotional feelings about an event. This paper explores the pre-processing stages of text classification for the emotion recognition based on Twitter conversations that correlate with the debate of Indonesian presidential candidates. Data pre-processing is an important step in sentiment analysis because the results of the analysis are strongly affected by the quality of the data provided. A combination of data processing has been carried out using Indonesian Twitter datasets. The accuracy of the analysis was tested using a deep learning model MLP and LSTM. The results show that the use of appropriate pre-processing techniques can improve accuracy.