Population-Scale Emotion Classification

NLP research · Freelance / University of Pennsylvania

I partnered with sociologists at the University of Pennsylvania on a long-term, population-level NLP study — building the model and the data pipeline behind it.

MillionsProfiles processed
ResearchSociology study

What I did

I built an emotion-classification model using then-current NLP and deep-learning methods, and designed Python scrapers to collect text from millions of public social-media profiles. Then I ran predictions and analyzed the results to measure population-wide shifts in emotional state following major political events.

What it demonstrates

Research-grade NLP applied at population scale, in service of a real academic question — and the end-to-end ownership of collecting the data, training the model, and turning predictions into findings a research team could use.