Content arrangement method, system, device and medium based on scene perception and feedback

By constructing a joint model of scene feature vectors and feedback feature vectors, an optimized playlist is generated, which solves the problem of difficult playback effect optimization in outdoor advertising screen systems and achieves efficient content arrangement and automated management.

CN122132595APending Publication Date: 2026-06-02BEIJING QUYUN WANWEI INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
BEIJING QUYUN WANWEI INFORMATION TECH CO LTD
Filing Date
2026-04-17
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing outdoor advertising screen control systems cannot comprehensively consider real-time scene factors and historical delivery feedback, resulting in difficulty in optimizing playback effects, high maintenance costs, and neglect of the combination relationship between advertisements and playback rhythm.

Method used

The current scene feature vector is constructed by multi-dimensional scene parameters, and feedback feature vector is constructed by combining historical content data. Joint modeling is performed to calculate the arrangement score, generate the target playlist, optimize the sorting through reinforcement learning, and introduce a feedback closed-loop correction mechanism for updating.

Benefits of technology

It improves the matching degree between advertising content and the placement scenario, reduces the proportion of invalid exposures, improves the overall advertising conversion efficiency, reduces manual maintenance costs, and achieves a higher degree of system automation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122132595A_ABST
    Figure CN122132595A_ABST
Patent Text Reader

Abstract

This specification discloses a content arrangement method, system, device, and medium based on scene perception and feedback. The method collects multi-dimensional scene parameters of the outdoor screen's location through various sensors and data platform interfaces, constructs a current scene feature vector, and filters a set of candidate content that meets the current scene from a content database based on the basic attributes of the content data and scene constraints. A feedback feature vector is constructed based on historical content data. The current scene feature vector and the feedback feature vector are jointly modeled to calculate the arrangement score of the content data in the current scene and sort it in descending order. A preset number of content data are combined to form an initial playlist. The initial playlist is then used as the optimization object for sorting and optimization to generate a target playlist. The content data is played on the outdoor screen according to the target playlist, and actual playback data is collected. The feedback feature vector and the parameters of the joint modeling are updated based on the actual playback data.
Need to check novelty before this filing date? Find Prior Art