Text moderation in Decision Engine

更新时间:
复制 MD 格式

The enhanced text moderation service can be integrated with Decision Engine. This integration allows you to pass Content Moderation detection results through Decision Engine as a pass-through service. This simplifies development and reduces integration costs. This topic describes how to use the text moderation service in Decision Engine.

Supported business scenarios

Decision Engine currently integrates two commonly used Content Moderation text moderation services: nickname_detection and comment_detection.

Service category

Business scenario

Scenario description

Decision Engine variable

Key value

nickname_detection

Applies to user nicknames, official account names, and live stream titles.

Focuses on detecting risks such as baseline violations, impersonation of official accounts, and prohibited traffic diversion to help govern fraudulent accounts.

enhanced text moderation - User nickname - Label

labels

enhanced text moderation - User nickname - fine-grained label

riskTips

enhanced text moderation - User nickname - Ad number matches

adNums

enhanced text moderation - User nickname - Matched risk words

riskWords

comment_detection

Applies to comments, bullet chats, public chat messages, and reposts.

Typically used for content with high risk rates, diverse risk types, and frequent variations. Detects risks such as baseline violations, ad-based traffic diversion, and prohibited content.

enhanced text moderation - Public chat and comments - Label

labels

enhanced text moderation - Public chat and comments - fine-grained label

riskTips

enhanced text moderation - Public chat and comments - Ad number matches

adNums

enhanced text moderation - Public chat and comments - Matched risk words

riskWords

Use cases

You can configure Content Moderation variables in Decision Engine to adjust the detection scope and the granularity of risk detection to meet your business requirements. You can also convert the labels returned by the Content Moderation service into entity-level (account, device, IP, etc.) risk profiles to manage entity-based risks more accurately and efficiently.

Use case

Description

Use the Content Moderation text moderation service in Decision Engine

If your application is already integrated with Decision Engine, you can perform content moderation directly through the engine, eliminating the need for a separate Content Moderation API integration.

Use the Content Moderation text moderation service in Decision Engine

You can adjust the scope and granularity of risk detection based on common industry practices or your specific business needs.

For example, if your business experiences frequent issues with ad-based traffic diversion, especially involving mobile numbers, you can configure a policy to flag text that matches these specific fine-grained labels.

Convert text-level risks to entity-level risk profiles

You can convert text-level risks into entity-level risks for accounts, devices, or IPs. This allows you to handle moderation results more accurately and efficiently.

For example, you can set a tolerance for the number of ad-related messages an account can send before a penalty is applied.

Related documents

For detailed instructions on the three scenarios, see Use the Content Moderation text moderation service in Decision Engine.