Public contracts hold crucial information: what governments purchase, how much they spend, and who the suppliers are. However, this data is often buried within large, unstructured, and incomplete document sets, making it difficult to analyze.
To tackle this challenge (before ChatGPT emerged), La Nacion, CLIP, Ojo Público, and MuckRock collaborated under the JournalismAI Collab. Together, we experimented with machine learning tools and techniques to develop a platform that helps investigative journalists process complex documents, uncover insights, and bring greater transparency to public spending.
My role
A key aspect of my work involved training the models to extract and structure key information from procurement documents, making it easier to identify patterns, detect irregularities, and cross-reference data across multiple sources. I also contributed to designing a user-friendly interface that allowed journalists to interact with the data intuitively, empowering them to conduct in-depth investigations without requiring advanced technical skills.