Product Overview

Update:

Multimedia Moderation

Multimodal content risk detection for web pages and documents

Multimedia Moderation is designed for web pages, H5 pages, forums, communities, e-commerce pages, blogs, campaign pages, PDFs, Word documents, spreadsheets, CSV files, and similar scenarios. It automatically detects risks in page and document text, images, links, QR codes, dynamic content, and contextual information, helping platforms quickly identify prohibited content, off-site traffic diversion, scam ads, and multimodal content risks.

Core Features

Feature Description
Web content parsing Automatically parses DOM, page text, images, comment areas, nicknames, buttons, overlays, and other page content, reducing the integration cost of extracting page content manually
Dynamic page rendering Supports JS-loaded content, H5, SPA, lazy loading, and pop-up content detection, covering more content visible to real users
Document content parsing Supports PDF, Word, Excel, and CSV parsing, extracting document text and embedded images for unified moderation
Image-text moderation Supports OCR, text extraction from images, and contextual semantic understanding to identify risks hidden in images, screenshots, and combined content
External link and QR code detection Detects redirect links, download links, social traffic diversion, QR code content, and risky domains to identify off-site diversion and fraud risks
Audio/video detection Detects audio and video content on websites with moderation models, reducing multimedia content violation risks

Moderation Workflow

URL/document input → Page rendering or document parsing → Content extraction → OCR → Text/Image/Audio/Video moderation → LLM semantic analysis → Risk aggregation → Result output

  1. Submit a web page URL or document URL. The system creates an asynchronous moderation task and returns a taskId.

  2. The service loads the page or downloads the document, and processes dynamic rendering, lazy loading, pop-ups, document type detection, and other content.

  3. The service parses DOM, body text, images, links, QR codes, document text, embedded images, and other content elements.

  4. Text, images, and OCR content are moderated, and contextual semantic analysis is applied.

  5. Multimodal results are aggregated into page-level or document-level risk conclusions and hit details.

  6. Get results by callback or the corresponding result query API.

Product Advantages

Advantage Description
High-coverage web parsing Covers static pages, dynamic H5 pages, SPA, comment areas, overlays, and lazy-loaded content to better restore the real page view
Multi-format document parsing Supports common office documents and spreadsheets, automatically extracting text and image content to reduce preprocessing work
Multimodal image-text understanding Combines text moderation, image moderation, OCR, and semantic understanding to improve risk detection on complex pages
Stable high-concurrency service The asynchronous task architecture is suitable for batch inspection, pre-launch review, and periodic review of existing pages
Flexible strategy configuration Supports different moderation strategies by business line, region, and content type, and uses strategyId for API calls

Use Cases

Use case Description
Social communities Moderate user profiles, post detail pages, comment aggregation pages, and topic pages
Game forums Moderate guides, forum posts, player profiles, and campaign discussion pages for violations and traffic diversion risks
E-commerce product pages Moderate product details, store profiles, buyer showcases, and marketing materials
AIGC platforms Moderate AI-generated articles, pages, image descriptions, and campaign landing pages
Document review Moderate text and image risks in user-uploaded PDFs, Word documents, spreadsheets, and CSV files
H5 campaign pages Moderate marketing campaigns, referral pages, lottery pages, and other pages for inducement, fraud, and sensitive content
Overseas content platforms Support multilingual risk inspection for overseas communities, blogs, forums, and content platforms

Integration

For web content, call the Web Moderation Task Submission API to submit a web page URL. For document content, call the Document Moderation Task Submission API to submit a document URL. After the API returns a taskId, get the result by callback or the corresponding result query API.