跨维度特征融合视角下的科技论文新颖性测量方法研究  

Measuring the Novelty of Scientific Papers Using Cross-dimensional Feature Integration

作　　者：马铭 郑哲浚 毛进[3,4] 白云[3] 李纲[4] Ma Ming;Zheng Zhejun;Mao Jin;Bai Yun;Li Gang(Research Institute for Data Management&Innovation,Nanjing University,Suzhou 215163;School of Information Management,Nanjing University,Nanjing 210023;School of Information Management,Wuhan University,Wuhan 430072;Center for Studies of Information Resources,Wuhan University,Wuhan 430072)

机构地区：[1]南京大学数据管理创新研究中心,苏州215163 [2]南京大学信息管理学院,南京210023 [3]武汉大学信息管理学院,武汉430072 [4]武汉大学信息资源研究中心,武汉430072

出　　处：《情报学报》2025年第3期339-352,共14页Journal of the China Society for Scientific and Technical Information

基　　金：国家自然科学基金面上项目“基于‘问题-方法’关联识别的科学知识创新探测与协同演化分析”(72174154);江苏省研究生科研与实践创新计划项目“双链融合视角下的战略新兴产业弱信号识别研究”(KYCX24_0315)。

摘　　要：准确评估科技论文的内在新颖性,对追踪学术前沿和实现高质量科研创新评价具有重要意义。本文基于科技论文的知识内容和学术交流结构,创新性地提出一种跨维度特征融合视角下的科技论文新颖性测量新方法,以全面、客观地评估科技论文的新颖性。首先,基于“问题-方法”组合,构建科技论文的结构化知识表示模型,并利用领域预训练语言模型为组合赋权;其次,考虑知识内容和学术交流结构,围绕原创性、复杂性和研究热度等事前特征,构建科技论文新颖性的跨维度综合测量指标;最后,通过对生物医学领域数据集的实证分析以及新颖性论文的事后影响力检验,证明了本文方法的有效性。实证分析结果表明,本文方法不受时间和环境因素干扰,能够在长时间跨度内保持方法效力并有效挖掘领域论文的新颖性模式。此外,与单一维度方法的对比证明了新方法能够更好地综合捕捉多维复合特征评估新颖性论文,避免了测量维度的单一化与片面化。本文为科技论文新颖性测度提供了新的视角和方法,同时可作为科研工作者识别和推广创新性研究的有效工具。Accurately assessing the intrinsic novelty of scientific papers is essential for advancing academic research and ensuring high-quality evaluation of scientific innovations.This study introduces a method for measuring the novelty of scientific papers that integrates cross-dimensional features based on their knowledge content and academic communication structure,enhancing the precision of academic assessments.First,a structured knowledge representation model for scientific papers is constructed using a specific combination of questions and methods,with a domain-pretrained language model employed to weigh these combinations.Second,from the perspectives of knowledge content and academic communication structure,we construct a cross-dimensional comprehensive measurement index to evaluate the novelty of scientific papers,focusing on ex ante features such as originality,complexity,and research popularity.The effectiveness of the proposed method is validated through empirical analysis of a biomedical dataset and ex post impact verification of novel papers.The empirical analysis results demonstrate that the proposed method is resilient to time and environmental factors,maintaining its effectiveness over long-term spans and successfully uncovering novelty patterns of papers in a specific field.Furthermore,comparisons with single-dimensional methods show that the proposed method synthesizes and captures multidimensional composite features more effectively,avoiding the oversimplification and one-sidedness of measurement dimensions.This study introduces new perspectives and methodologies for measuring the novelty of scientific papers and offers researchers a valuable tool for identifying and advancing innovative research.

关 键 词：新颖性测量 跨维度特征融合 “问题-方法”组合 事前特征 事后影响力 

分 类 号：G250[文化科学—图书馆学] G250.2