GDC 2025 Session Viewer
Machine Learning Summit: Twitch Chat Safety: Scalable and Personalized Moderation with Deep Learning
Linda Liu (Senior Applied Sceintist, Twitch)
Pass Type: All Access Pass, Summits Pass - Get your pass now!
Track: Programming
Format: Lecture
Vault Recording: TBD
Audience Level: Intermediate
Twitch's mission is to create belonging by enabling streamers to build community. To help provide a safe environment for streamers, Twitch developed automated moderation for chat. Currently powering over 95% of Twitch chat globally, Smart Detection employs scalable NLP models that can learn individual channels' moderation preferences. This session explores the development and evolution of Smart Detection, including the model development cycles, and lessons learned from production. Attendees will gain insights into the application of advanced AI and methods to leverage LLM for labeling. This talk provides practical knowledge and inspiration for developing and scaling automated moderation tools.
Takeaway
Attendees will learn about the latest developments in Twitch chat content moderation and gain practical insights into the application of deep learning and generative AI.
Intended Audience
This talk is for anyone passionate about content moderation and Natural Language Processing (NLP).
