Note content promotion and marketing optimization method based on large language model
By combining large language models and the MAP-Elites algorithm, we generate and optimize creative content for note promotion, solving the problems of material homogenization and fluctuating campaign performance, and achieving broad coverage and efficient iterative marketing optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU CHUXIN CULTURE MEDIA TECHNOLOGY CO LTD
- Filing Date
- 2026-02-28
- Publication Date
- 2026-05-26
AI Technical Summary
Existing large language models suffer from issues such as homogenized materials, fluctuating campaign performance, difficulty in simultaneously ensuring diversity and quality stability, and difficulty in adaptively adjusting the creative library when the campaign cycle changes.
By using a large language model to encode note data and delivery context data, creative semantic vectors and filter values are generated. Cell sets are formed through coverage decomposition and clustering. A cell elite library is established by combining the MAP-Elites algorithm. Transfer mutation is performed to generate candidate texts. The elite library is updated based on quality indicators to achieve closed-loop optimization of creative search and delivery evaluation.
It achieved optimized results with broad creative coverage, high iteration efficiency, and strong adaptability to different campaigns, reducing the homogenization of creative materials and improving the stability and exploration efficiency of the campaign process.
Smart Images
Figure CN122087347A_ABST