基于大语言模型的论文创新质量测度研究  

Research on Innovation Quality Measurement of Papers Based on Large Language Model

作　　者：李晶[1] 邱昕鹏 Li Jing;Qiu Xinpeng(School of Information Management,Sun Yat-sen University,Guangdong Guangzhou 510006)

机构地区：[1]中山大学信息管理学院,广东广州510006

出　　处：《情报理论与实践》2025年第3期169-177,共9页Information Studies:Theory & Application

基　　金：国家社会科学基金项目“科技论文创新质量的微观测度及应用研究”的成果,项目编号:22BTQ097。

摘　　要：[目的/意义]在破“五唯”背景下,对学术论文创新质量的准确测度将为完善科技成果评价体系提供方法基础。[方法/过程]提出一种综合考虑论文新颖性与有用性的论文创新质量测度模型,同时创新性地应用LLaMA3-BPE融合模型改进传统的文本关键词抽取技术,提升新颖性关键词抽取的有效性。[结果/结论]提出的创新质量测度模型在一定程度上解决了现有论文创新性评价存在的低区分度、评价维度不全等问题,LLaMA3-BPE融合模型的应用能有效改进文本关键词抽取的准确性。[Purpose/significance]Under the background of breaking the“five principles”,the accurate measurement of the innovation quality of academic papers will provide an methodological basis on the evaluation system of scientific and technological achievements.[Method/process]This paper proposes a paper innovation quality measurement model that comprehensively considers the novelty and usefulness of the paper,and innovatively applies the LLaMA3-BPE fusion model to improve the traditional text keyword extraction technology,and improve the effectiveness of the novelty keyword extraction.[Result/conclusion]The innovation quality measurement model proposed in this paper solves the problems of low discrimination and incomplete evaluation dimensions existing in the innovation evaluation of existing papers to a certain extent,and the application of LLaMA3-BPE fusion model can effectively improve the accuracy of text keyword extraction.

关 键 词：论文评价 创新质量 测度模型 大语言模型 实体抽取 

分 类 号：G301[文化科学] G353.1