Tweet Sentiment Analysis at Booz Allen Hamilton
As a Summer Games Intern at Booz Allen Hamilton (May–August 2019), I analyzed public attitude toward police across South American cities significantly affected by gang violence, using Twitter as the data source. The question was practical: where does sentiment toward local authorities differ within South America, and how do you show that clearly enough for decision-makers to act on it?
What I did
- Ran sentiment analysis on thousands of tweets that mentioned local authorities and originated in the affected cities.
- Built interactive Tableau visualizations so results could be compared geographically. The deliverable was a view a decision-maker could explore, not a notebook of scores — the value was seeing where attitudes differ from place to place, not any single sentiment number.
The sentiment analysis produced the signal; the Tableau work made it legible to people who were never going to read raw tweet data. The broader internship also included Python data ingestion and cleaning for a federal-client prototype, PostgreSQL-backed workflows, a Spring Boot API concept, and documentation and test cases for handoff.
What it taught me
Tweets are messy input, and this project forced me to understand what the data actually contained before trusting any aggregate built on top of it. It also taught me that building for a non-technical audience is its own engineering constraint — an interactive view a stakeholder can explore is a fundamentally different deliverable than an analyst's results file.
This was client work, so there is no public code, data, or dashboard to share.