Using text moderation
PlayFab Party offers a text moderation solution that is integrated with text chat. Text chat can be moderated in real-time to filter out offensive language. This feature is backed by Azure AI Content Safety. For more information, see Azure AI Content Safety.
Language support
Text moderation is supported for more than 100 languages. However, you don't need to do anything to configure a language selection for text moderation; Azure automatically detects the language of each text message. For a complete list of supported languages that support autodetection and profanity filtering, refer to Azure AI Content Safety's list of supported languages.
Enabling text moderation
Text moderation is configured per chat control. It's enabled by calling PartyLocalChatControl::SetTextChatOptions() with the PartyTextChatOptions::FilterOffensiveText option. Chat controls that have moderation enabled receive text chat with offensive words masked by asterisks. Since moderation is enabled per chat control, you're able to create UI to allow users to set their own moderation choice. Alternatively, you can enforce moderation for some or all users.
To disable moderation for a chat control, use the same function, passing in a PartyTextChatOptions option that doesn't contain the FilterOffensiveText flag.
Sample code:
// Opt into moderation
myLocalChatControl->SetTextChatOptions(PartyTextChatOptions::FilterOffensiveText, nullptr);
// Opt out of moderation
myLocalChatControl->SetTextChatOptions(PartyTextChatOptions::None, nullptr);
Sending and receiving moderated text
All text sent to a chat control with moderation enabled has moderation applied, regardless of the sender's own moderation settings. Text messages received by a chat control with moderation enabled has the chatText field in PartyChatTextReceivedStateChange replaced with the moderated version.
More fields are present on PartyChatTextReceivedStateChange to enable more complex scenarios, such as giving users an option to view an unfiltered version of the text. For more information, see the PartyChatTextReceivedStateChange reference page.
If moderating a text message fails, either due to a service error or because the text was deemed offensive by the service without being able to identify specific terms, then the entire text message is masked with asterisks. The PartyChatTextReceivedOptions enum provided in the options field of PartyChatTextReceivedStateChange contains more information when this error case occurs.
Adjusting filtering levels
For English, the level of filtering that happens for a chat control can be adjusted by using the PartyManager::SetOption function. The filtering levels available are Family-Friendly, Medium, and Mature. The default level is Family-Friendly.
Sample code:
// Set the filter level to Mature
PartyTextChatFilterLevel filterLevel = PartyTextChatFilterLevel::Mature;
PartyOption option = PartyOption::TextChatFilterLevel;
PartyError error = PartyManager::SetOption(nullptr, option, &filterLevel);