Social Media Emotion Analysis in Indonesian Using Fine-Tuning BERT Model

Social media is the leading platform were users' express opinions and emotions. Emotion Analysis aims to identify emotions: happy, sad, angry, fear, disgust, shame, and guilt. InaMoodMeter framework takes the unstructured status of a Facebook user. It processes it to extract emotion from teenage users for self-assessment and to observe emotion related to online customer satisfaction. We also created an Indonesian Dataset for Emotion Analysis from Facebook. The dataset can be utilized as a valuable benchmark for emotion classification in Indonesian. With this dataset, many state-of-the-art approaches are evaluated. We also experimented on the ISEAR Indonesian translated dataset. This research consists of two stages: obtaining training data to build the dataset and performing classification. Our proposed method can achieve the highest accuracy of 79% with the Fine-Tuning Bert Model.