A mechanism is proposed for identifying policy violating videos based on user behavior. A content item service may identify a first video associated with a first probability value for the first video of violating one or more policies set within a content item service platform. The content item service may identify a plurality of other videos that have been co-watched with the first video using a co-watch graph. The content item service may calculate a score for the first video based on probability values for the plurality of other videos of violating the one or more policies and likelihood values for the plurality of other videos of being watched with the first video. The content item service may determine whether to initiate remedial actions regarding the first video based on the score.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Lane, Corey; Long, Eileen; Chambers, William; and Krovblit, Guillermo, "IDENTIFYING POLICY VIOLATING VIDEOS BASED ON USER BEHAVIOR", Technical Disclosure Commons, (November 14, 2018)