A method for identifying change based on content-based embedding is disclosed. The proposed method computes content similarity of content items uploaded to a user’s personal channel. The method then computes a channel risk score, which identifies whether the type(s) of content uploaded to a user’s personal channel has changed after a specific point in time. A channel in which the type(s) of content has changed after a specific point in time, for example after the channel was reviewed for a monetization program, may indicate potential abuse of the monetization program. Based on the risk score, the method may flag potentially abusive channels for administrative review, which may be reviewed by an automated system, a human reviewer, or a combination.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.