Overview of Content Moderation Capabilities

The Instant Messaging (IM) service supports content moderation, offering basic moderation capabilities, professional moderation services, and the ability to integrate third-party moderation services.

• Built-in Sensitive Word Filtering: The IM service comes with an out-of-the-box sensitive word filtering mechanism. Note that this mechanism is a basic protection feature and is limited to text content (with a default maximum of 50 sensitive words). It is not a substitute for professional content moderation services.
• Pre-messaging Callback Service: The Pre-messaging callback service allows messages to be first sent to an address specified by the app, where the app can decide whether to proceed with delivery to the recipient. Apps can use the Pre-messaging Callback Service to implement their own moderation or integrate third-party moderation services.

If a message is flagged as violating content guidelines and cannot be delivered to the recipient, the sender will not be notified by default. If you wish to enable notifications for senders, here’s how:

• Client-Side: Supports notifying the sender. If the app wants the sender to receive a notification that the message has been blocked, you can enable the Blocked Message Status Callback to Sender feature in the RC Console under the Basic Features section and set up a listener on the client-side (requires Android/iOS SDK version ≧ 5.1.4, Web ≧ 5.0.2). For details, refer to the "Message Block Notification Callback" documentation for each client.
• Server-Side: Supports providing Audit Result Callback for messages that trigger the Professional IM Moderation Service. Both approved and rejected messages will be returned through this service.

Sensitive Word Filtering Mechanism

The Sensitive Word Filtering Mechanism is a basic protection feature that only supports the identification and filtering of sensitive words in text message content. For messages that contain sensitive words, you can choose to block the message (preventing it from being delivered to the recipient) or replace the sensitive words according to specified rules before delivery.

Currently supported languages for sensitive word filtering include: Chinese, English, Japanese, German, Russian, Korean, and Arabic.

tip

For moderating text (with semantic detection), images, audio clips, and short videos, it is recommended to use the Content Moderation Service provided by the IM service.

You can manage sensitive words for the development or production environment under your App Key via the Console or API:

• In the Console, under the Security & Moderation section, the Sensitive word Settings page supports adding, deleting, and exporting sensitive words. For details, refer to Configuring Sensitive Word Service.
• The Server API supports adding, deleting, and retrieving sensitive words. For details, refer to Message Sensitive Words.

Default Behavior

• By default, a maximum of 50 sensitive words can be set.
• By default, it only applies to messages sent from the client SDK.
• By default, it only supports identifying sensitive words in the official built-in text message type (message identifier RC:TxtMsg). Supported conversations include one-to-one chat, group chat, chatroom, and ultra group. Modified text messages in ultra groups are also subject to sensitive word identification, blocking, or filtering by default.
• By default, there is no notification when sensitive words are triggered.

Sensitive Word Filtering Rules

• Simplified and Traditional Chinese Smart Filtering: After setting a simplified Chinese sensitive word, the corresponding traditional Chinese sensitive word will also be automatically identified and filtered.
• Smart Ignoring Punctuation: Supports intelligently ignoring punctuation in strings to match sensitive words. For example: If the sensitive word is “反动”, messages containing “反，动” will also be filtered. In this case, “反v动” and “反3动” will also be filtered.
• Smart Recognition of English Words: Supports intelligent recognition and filtering of English words to improve matching accuracy. English supports automatic matching of full-width, half-width, uppercase, and lowercase. For example: If the sensitive word is “AV”, only the complete words “AV” and "av" will be filtered, while the “av” in the word “Java” will not be filtered.
• Numeric Sensitive Words: Supports matching continuous numeric strings and intelligently matching full-width numbers. For example: If the sensitive word is “123”, it can filter “123”, “1 2 3”, “123汉字”, and “１２３” (full-width), but it cannot filter the “123” segment in “123467”.

Pre-messaging Callback Service

If you want to integrate your own moderation system or other third-party content moderation services, you can use the Pre-messaging Callback Service.

The Pre-messaging Callback Service (formerly Template Routing) provides a message filtering mechanism. You can synchronize messages to your specified server based on parameters such as sender ID, recipient ID, message type, and conversation type. In ultra group conversations, modifications to message content and message extensions are also supported through the Pre-messaging Callback service.

After messages are synchronized to your specified server, you can use your own moderation system to perform content moderation or integrate other third-party moderation systems. The IM service will decide whether to deliver the message, replace the content, and how to replace it based on the response returned by your app server.

You can manage the status and routing rules of the Pre-messaging Callback Service for the development or production environment under your App Key in the Pre-messaging Callback Service page of the Console.

For details on how to create routing rules and the specific parameters of the callback, refer to the Pre-messaging Callback Service documentation.

Pre-messaging Callback Service Pricing

Pricing is based on the Pricing Documentation.