Abstract extraction and model training methods, equipment and media
By identifying office scenarios and building an adaptive summary extraction model, the problem of poor model generalization in existing technologies is solved, enabling diverse summary extraction in different office scenarios and improving office efficiency.
CN122309727APending Publication Date: 2026-06-30SF TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SF TECH CO LTD
- Filing Date
- 2024-12-30
- Publication Date
- 2026-06-30
AI Technical Summary
Technical Problem
In existing technologies, models trained specifically for a particular office scenario cannot be directly applied to other scenarios, exhibiting poor generalization and generating abstracts that are formulaic and lack diversity.
Method used
By identifying the office scenarios of unread target data, a summary extraction model is used to extract summaries based on these scenarios, and a scenario-adaptive summary extraction model is constructed. Through fine-tuning of a large language model, summaries suitable for different office scenarios are generated.
Benefits of technology
It enables scenario-adaptive summary extraction of unread data in different office settings, ensuring the diversity and accuracy of summaries and improving office efficiency.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122309727A_ABST
Abstract
This application provides a summary extraction and model training method, device, and medium. Upon identifying unread data (target unread data) in any office scenario, the method determines the target office scenario, i.e., the office scenario to which the target unread data belongs. A summary extraction model is then used to extract summaries of the target unread data based on the target office scenario, yielding a summary of the target unread data. The target summary model is used to extract summaries of unread data for different office scenarios. Thus, the summary extraction model enables scenario-adaptive summary extraction, ensuring the diversity of the extracted summaries.
Need to check novelty before this filing date? Find Prior Art