Project overview
Natural language processing over a large drug-review dataset, surfacing sentiment and safety signals from patient-written text.
Drug review sites hold thousands of first-hand accounts of what a medication actually did to someone, and almost nobody reads them past the star rating. RxSentinel treats that written text as data. It scores every review for sentiment, groups the results by drug, and puts the whole thing behind a dashboard where one medication can be held up against another rather than read one review at a time.
The star rating and the text often disagree, so the app makes that visible: a correlation heatmap sets rating against sentiment score and usefulness, and word clouds surface the terms patients keep reaching for, which is usually where side effects show up. A Prophet forecast projects the review timeline forward, so a drug starting to collect complaints is visible before the average moves. Findings export to PDF, so something noticed here can leave the app.