Multi-mode sensitive tag identification method and system for short video content security

By constructing a logical chain of content associations and weaving features from multi-dimensional representation units, the problems of low efficiency and accuracy in the security review of short video content are solved, and efficient and accurate sensitive tag identification and tag relationship presentation are achieved.

CN122223610APending Publication Date: 2026-06-16GUANGZHOU AVANTI ELECTRONIC TECH CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202610304522.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-13
Publication Date
2026-06-16

AI Technical Summary

Technical Problem

Existing technologies are inefficient, costly, and prone to missed or false detections in short video content security review, and cannot comprehensively and accurately identify multimodal sensitive content.

Method used

By anchoring the sensitive representation subjects of short videos, we decompose them into multi-dimensional fine-grained representation units, mine the logical chain of content association, construct a network of sensitive association features, perform dynamic anchoring and matching of tag nodes and reverse source tracing calibration, and extract and aggregate sensitive tags.

Benefits of technology

It has significantly improved the accuracy and efficiency of sensitive tag identification, clearly presented the relationships and hierarchical structure between tags, and effectively maintained the health and security of the network environment.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a multi-modal sensitive label identification method and system for short video content security, and the method comprises the steps: anchoring a short video sensitive representation main body, carrying out the disassembly of the main body to form multi-dimensional fine-grained representation units, mining a content association logic chain between the multi-dimensional fine-grained representation units, and obtaining a related set; executing cross-dimension feature interleaving and link fusion processing to construct a sensitive association feature weaving network; label node dynamic anchoring matching is carried out on the sensitive association feature knitmesh and a preset sensitive label entry set, and an initial anchoring association relationship is generated; feature network reverse traceability and label anchoring feedback calibration processing is carried out on the initial anchoring set, and anchoring correlation deviation is corrected; and extracting sensitive tags based on the calibrated tag feature node anchoring set, executing association aggregation and hierarchical division processing among the tags, and outputting a sensitive tag set corresponding to the short video. According to the method, the multi-mode information of the short video is comprehensively utilized, and the accuracy of sensitive tag identification is improved.
Need to check novelty before this filing date? Find Prior Art