Tisane is widely used to moderate live communications between users on group instant messaging platforms. Usually, in live chats Tisane runs in fully automatic mode.
If false positives is a sensitive issue, then an appeal procedure and human review (on request) can be established.
As shown on the diagram below: a simple integration architecture scans every message and blacklists or sends the message to a special channel if it is deemed problematic and the policies do not allow the type of abuse that was detected.
- The client application sends posts to Tisane for scanning.
- Tisane flags posts based on severity level and abuse type.
- The client application then must check if the types of entries logged in the
abusesection are not allowed by the community policies. - If there are restricted types, then the post is blacklisted. (Presumably, the post is sent to a special channel for moderators. For maintenance reasons, it is not recommended to discard messages without a trace.)
- If there are no restricted types or there is no
abusesection in Tisane's response, the post is published.
Our partner PubNub built a content moderation demo seamlessly integrated with Tisane and published its source code in GitHub. The implementation allows the users dynamically set community moderation policies.