Project

Sentiment Drift and Bot-Like Text Detection

Python Machine Learning RoBERTa XGBoost

Machine learning project visual

This project focused on tracking how public sentiment changes over time while accounting for bot-like text that can distort the signal. The pipeline collected Bluesky posts, cleaned and filtered the text, classified sentiment, and produced time-series visualizations that made sentiment drift easier to inspect.

My work centered on the machine learning and analysis workflow. I helped fine-tune a RoBERTa sentiment model, engineered TF-IDF and XGBoost-based bot detection, and connected those outputs to visualizations that showed how public opinion shifted across the collected data.

The project strengthened my experience with practical machine learning systems: data collection, text preprocessing, model tuning, classification, and presenting results in a way that helps people reason about the data rather than just consume a raw score.

Source: BryanNak/SentimentAnalysis