For a streamer donation platform: a machine-learning system that reads context and patterns to catch illegal gambling content that keyword blocking missed.
Bagi Bagi is a donation platform that lets fans support streamers financially. The problem they faced was illegal gambling content slipping through content moderation. Standard keyword-blocking wasn't catching it because bad actors changed their wording constantly. We built a detection system that reads context and communication patterns instead, using machine learning to recognize gambling-related activity even when the specific words aren't there. It adapts over time rather than requiring manual updates to block lists.
What it changed
01Detects nuanced patterns, not just banned keywords
02Adapts to new phrasing without manual rule updates
