论文学术创新评价是科技成果创新评价的重要组成部分，对论文进行创 新评价研究有助于提升评审效率、推动成果转化。针对目前科技成果多元价值 评价需求和论文内容细粒度挖掘趋势，结合知识实体抽取与引文评价句分析， 本研究对论文的创新程度做定量化评价研究。
论文的研究内容在与已有研究的对比中体现出其新颖性，而论文经引用 体现其研究内容的创新价值，因此，结合上述两个过程，可以量化论文创新性， 为论文创新评价研究提供新思路。首先，通过论文内容细粒度挖掘，抽取其研 究问题实体和研究方法实体，用问题实体与方法实体组合的形式代表论文研 究内容，采用词频和相似度结合的方式计算原文新颖性。其次，将引用功能分 为问题型引用和方法型引用，通过不同类型的引用验证论文在研究问题和研 究方法两方面产生的价值，并衡量其价值比重，为不同论文动态分配新颖性计 算结果的权重，完成对论文创新价值的精细化度量。
对 CNKI 中收录的“信息抽取”研究主题下的共计 3597 篇中文期刊论文 进行实证研究，经实例分析检验该方法有效。结果表明，本研究提出的创新性 测度方法能从原文与引文两个角度充分揭示论文创新成果，能够在细粒度层 面更充分地挖掘论文多元价值。

The academic innovation evaluation of papers is an important part of the innovation evaluation of scientific and technological achievements, and the innovation evaluation research of papers is conducive to improving the review efficiency and promoting the transformation of achievements. Aiming at the current needs for multiple value evaluation of scientific and technological achievements and the trend of fine-grained mining of paper content, combined with the extraction of knowledge entities and the analysis of citation evaluation sentences, this study conducts quantitative evaluation research on the innovation degree of papers.
The research content of the paper reflects its novelty in the comparison with the existing research, and the paper reflects the innovative value of its research content through the citation process, so combining the above two processes can quantify the innovation of the paper and provide new ideas for the innovation evaluation of the paper. Firstly, through the fine-grained mining of the content of the paper, the research problem entity and the research method entity are extracted, the research content of the paper is represented in the form of the combination of problem entity and method entity, and the novelty of the original text is calculated by combining word frequency and similarity. Secondly, the citation function is divided into problem-based citation and methodological citation, and the value generated by the paper in both research questions and research methods is verified by different types of citations, and its value proportion is measured, and the weight of the novelty calculation results is dynamically assigned to different papers, so as to complete the refined measurement of the innovation value of the paper.
A total of 3597 Chinese journal papers under the research theme of "information extraction" included in CNKI were empirically studied, and the effectiveness of this method was verified by case analysis. The results show that the innovative measurement method proposed in this study can fully reveal the innovative results of the paper from the perspectives of the original text and the citation, and can fully explore the multiple values of the paper at the fine-grained level.