Customer Review Sentiment Analysis (NLP)
Cleaned and sentiment-scored 10,000 Yelp reviews with VADER, found lexicon-based scoring misclassified 44.9% of 1-star reviews as positive due to sarcasm, and corrected 1,020 reviews by combining star rating with sentiment score; keyword/bigram analysis identified service quality, order accuracy, and wait time as the top complaint drivers. In progress: restructuring into a script-based ETL pipeline reading/writing from AWS S3, and building the SQL and Power BI layers.
