| 摘 要: 针对现有文本生成类多智能体协同框架存在的引用文献置信度低、议题论点偏移等问题,提出一种科研论文写作自动化框架。在知识构建层面,通过多向数据源融合机制整合网络搜索引擎与大模型文献检索技术,构建具备高置信度的知识图谱;在协同生成层面,将低频次知识再唤醒机制纳入桥接优化器,支撑去中心化与并行化结合的混合式多智能体协同架构,强化论文议题内容的覆盖面与聚焦度。在3个不同主题的论文生成任务下,基于量化与人工评估进行二维测量,所提框架在文章覆盖面、聚焦度与引用文献置信度方面均优于对比技术框架。关键词:科研论文写作;多向数据源融合;知识图谱;桥接优化器;多智能体协同 |
| 关键词: 科研论文写作 多向数据源融合 知识图谱 桥接优化器 多智能体协同 |
|
中图分类号:
文献标识码: A
|
|
| Research on Scientific Paper Writing Based on Trusted Knowledge Graph and Multi-Agent Collaboration |
|
LIU Yichen, CHEN Yunfang, QI Dazhi, ZHANG Wei
|
(School of Computer Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, China)
929779718@qq.com; chenyf@njupt.edu.cn; 1552817181@qq.com; zhangw@njupt.edu.cn
|
| Abstract: Aiming at the issues of low credibility of cited references and topic drift in existing mult-i agent collaboration frameworks for text generation, this study proposes an automated scientific paper writing framework. At the knowledge construction level, a mult-i directional data source fusion mechanism integrates web search engines and large language mode-l based literature retrieval to build a high-credibility knowledge graph. For collaborative generation, a low-frequency knowledge reactivation mechanism is incorporated into a bridge optimizer, supporting a hybrid mult-i agent collaboration system that combines decentralization and parallelism. This enhances both the coverage and focus of the paper’s content. Under three paper generation tasks with different themes, a two-dimensional evaluation consisting of quantitative metrics and manual assessment is conducted. Experimental results show that the proposed framework achieves better performance than comparative frameworks in terms of content coverage, topic focus and credibility of cited references |
| Keywords: scientific paper writing multi directional data source knowledge graph bridge optimizer multiagent collaboration |
