Introduction to Content Moderation

更新时间:
复制 MD 格式

Content Moderation is built on Alibaba Cloud's visual analysis and deep learning technologies. The service is widely used and continuously optimized across various realms and scenarios within Alibaba Group and by cloud customers. It provides comprehensive services for threat detection and administration, including image recognition, localization, and retrieval. Content Moderation helps reduce compliance threats from pornography, terrorism, advertisements, junk content, and sensitive information. It also significantly reduces manual review costs.

Activate the service

To activate the service, click Activate Now on the product page. For more information, see Quick start.

Features

The Alibaba Cloud Visual Intelligence API provides the following Content Moderation features:
CapabilityFunctionDescription
Image Content ModerationIntelligent pornography detectionDetects pornographic and vulgar content in images.
Sensitive content detectionDetects content in images related to sensitive events, violence, weapons, terrorism, gore, and explosions.
Logo detectionDetects content such as TV station logos, watermarks, and trademarks in images.
Junk ad detectionDetects advertising information in images. It is especially effective at detecting text-based ads.
Undesirable scene detectionDetects content in images related to drugs, gambling, and Picture-in-Picture (PiP).
High-risk person detectionDetects sensitive individuals and celebrities in images.
Text Content ModerationJunk content detectionDetects junk content in text by combining behavior and content analysis. It uses multiple dimensions, models, and detection methods. This feature is suitable for detecting junk text in images.
Ad content detectionDetects ad content in text by combining behavior and content analysis. It uses multiple dimensions, models, and detection methods. This feature is suitable for detecting ad content in images.
Sensitive content detectionDetects politically sensitive content in text by combining behavior and content analysis. It uses multiple dimensions, models, and detection methods. This feature is suitable for detecting sensitive text in images.
Terrorism content detectionDetects terrorism-related content in text by combining behavior and content analysis. It uses multiple dimensions, models, and detection methods. This feature is suitable for detecting terrorism-related text in images.
Abuse detectionDetects abusive content in text by combining behavior and content analysis. It uses multiple dimensions, models, and detection methods. This feature is suitable for detecting abusive text in images.
Pornography detectionDetects pornographic content in text by combining behavior and content analysis. It uses multiple dimensions, models, and detection methods. This feature is suitable for detecting pornographic text in images.
Spam and flood detectionDetects spam and flood content in text by combining behavior and content analysis. It uses multiple dimensions, models, and detection methods. This feature is suitable for detecting spam and flood text in images.
Prohibited content detectionDetects prohibited content in text by combining behavior and content analysis. It uses multiple dimensions, models, and detection methods. This feature is suitable for detecting prohibited text in images.

Scenarios

Content Moderation can be used in the following scenarios:
  • New Retail

    This solution provides multi-dimensional threat monitoring for multimedia content in New Retail scenarios. This includes product compliance review, moderation of promotional content, monitoring of comments and messages, ad content moderation, and product infringement detection. For example, the solution can detect pornography, vulgarity, gore, violence, and prohibited content in product images. It can also detect spam in product descriptions, identify spam and prohibited items in comments, and detect ads and logos in product images.

  • Media

    This solution addresses common issues in the media industry. It prevents the spread of content such as images that insult martyrs, parody leaders, or contain illegal or politically sensitive material. The algorithms in Content Moderation can accurately and efficiently detect sensitive individuals and text within images.

  • Smart Security

    This solution provides multi-dimensional threat monitoring for offline multimedia content in the security industry. It covers security related to people, objects, and scenario-based behaviors. Examples include facial recognition for access control, facial monitoring of key individuals at surveillance checkpoints, crowd gathering detection, detection of events that may affect public stability, and fire safety monitoring (such as detecting smoking, sparks, fire, and explosions).

For more product updates, follow the Alibaba Cloud Visual Intelligence API.