Stop Fraud Ads Reporting Support
AI-powered classification, screening, and discovery support for the Digital Democracy 2030 fraud-ad reporting site.
Technologies Used
Key Features
Overview
As part of Digital Democracy 2030, this project released a beta version of a fraud-ad reporting site designed to turn citizen reports into actionable signals.
Role
Naoki worked on backend features using generative AI. The system classifies reported ads and posts, labels categories such as impersonation and investment scams, extracts public figures and summaries, and supports routing reports to appropriate review organizations.
Technical Focus
The project introduced a self-improving mechanism that uses past labeling results to improve future classification accuracy. For impersonation images, prompt tuning and data processing were optimized to improve extraction and judgment quality.