Guide:
- Final paper: analysis and results of the project
- ipynb file: code for the entire project
- Powerpoint: project proposal and initial insights
Sentiment Analysis of Donald Trump’s Tweets During Key Political Events: Comparing RoBERTa and LSTM Models
Authors: Cecily Wang-Munoz and Cole Whittington
Social media plays a key role in shaping public narratives, particularly when used by political leaders to communicate during key events. President-elect Donald Trump was particularly unique because he was the first president to use Twitter as a major form of communication with the American public. This project aims to analyze the sentiment of tweets authored by Trump, focusing on how his messaging shifted throughout his presidency and how the sentiment behind his tweets changed engagement from Twitter users. By comparing the performance of RoBERTa, a transformer-based Natural Language Processing model, with an LSTM neural network, we investigate which approach better captures sentiment shifts and emotional tones in his messaging.
Donald Trump’s tweets provide a unique case study, reflecting his use of social media as a direct communication tool, a method he uses in a robust manner. As a high-profile individual for many years, Trump has long used his influence to express his opinion about politics and his career. Analyzing these tweets helps uncover temporal patterns in sentiment and emotion, showing how his language evolved in response to events or initiated them. At the very end of his presidency, on January 6, 2021, his Twitter influence was even further put on display after he was impeached by the House of Representatives for inciting an insurrection. Soon after, on January 8, Twitter banned Trump’s account, which was reinstated by then-new CEO Elon Musk nearly two years later, in November 2022.