TEXT EMOTION ANALYSIS USING DEEP LEARNING ON NATIONAL ELECTION

Indonesia's digital landscape, notably social media, is increasingly influencing politics, necessitating sophisticated analysis methods to comprehend public sentiment. However, traditional sentiment categorization is often biased, underscoring the need for nuanced emotion analysis. This study addresses this by using deep learning algorithms for emotion analysis on national elections data, aiming to accurately understand public emotions. This study began with a complex data collection phase, using manual downloading and data crawling techniques to gather tweets from Twitter. This was followed by extensive preprocessing, which involved case folding, cleansing, tokenizing, handling slang, and stemming to prepare the data for further stages. The refined data was split for modeling and analysis, with the BERT algorithm applied for training and subsequent evaluation. Ultimately, the most accurate model was used for emotion recognition on the election dataset, facilitating a comprehensive data exploration and analysis. The study resulted in a proficient use of the BERT model for emotion classification, revealing its capability for comprehensive emotion detection in varied text data. Additionally, the emotion analysis executed on an election-related Twitter dataset exhibited a polarity in the emotional landscape, with strong emotions such as "Anger" and "Love" dominating, thus mirroring diverse public opinions towards election events.

Keyword: BERT, Emotion Analysis, National Election, Social Media