基于一致性判别的可控新闻稿件生成方法、设备及介质
By constructing news generation and consistency discrimination datasets, training generation and discrimination models, and combining them with a scoring function, the consistency problem of the generation model in news article generation was solved, achieving controllable news article generation and improving the usability of the generated text.
CN117786095BActive Publication Date: 2026-07-17CHENGDU SOBEY DIGITAL TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU SOBEY DIGITAL TECH CO LTD
- Filing Date
- 2023-12-14
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing generation models struggle to guarantee consistency in format, style, and factual accuracy during press release generation, increasing the workload for editors.
Method used
Construct a news generation dataset and a consistency discrimination dataset, train the news generation model and the consistency discrimination model, and combine the beam search and the scoring function of the consistency discrimination model to generate news articles that meet user needs.
Benefits of technology
It enables controllable news article generation based on user needs, improves the usability of generated text, and reduces the subsequent workload of editors.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN117786095B_ABST
Abstract
本发明公开了一种基于一致性判别的可控新闻稿件生成方法、设备及介质，属于新闻文本生成领域，包括步骤：S1，构造新闻生成数据集并基于所构造的数据集训练新闻生成模型；S2，构造一致性判别数据集并基于所构造的数据集训练一致性判别模型；S3，结合步骤S1、S2中训练的新闻生成模型和一致性判别模型构造基于一致性判别的可控新闻稿件生成模型进行新闻稿件的可控生成。本发明能够高效整合新闻资源、可控地生成满足用户需求的新闻稿件，从而提高所生成的新闻稿件的可用性，减少用户后续工作量。
Need to check novelty before this filing date? Find Prior Art