This project tackles the notoriously difficult problem of predicting which of the thousands of bills introduced in Congress will actually become law, creating genuine value for advocacy groups, policy researchers, and government affairs professionals who currently rely on intuition and limited data. We’ll build an end-to-end pipeline that
(1) scrapes data from various online APIs about congressional bills, including the text, the sponsors and cosponsors, the amount of lobbying activity behind the bill, social media trends, news article mentions, etc..
(2) uses NLP techniques and LLMs (GPT-4/Claude) to extract complex features like bill complexity scores, stakeholder impact analysis, and policy domain classifications from legislative text—features that traditional political science models might not capture entirely,
(3) employs different classification and regression methods to predict (a) whether a bill passes both houses of Congress, and (b) how long it might take for the bill to pass (using survival modeling tools).