编辑出版中心知识库

摘要：［目的/意义］创新性是对学术论文质量最基本的要求，是学术论文的灵魂，是学术论文评价的核心。知识元是学术论文基本组成单元。基于知识元理论和机器学习相关理论与算法，从学术论文内容层面研究计算机如何智能化地进行创新性评价及其实现过程与方法。［方法/过程］首先，构建学术论文的研究问题、理论、方法、结论4 个知识元本体，接着提出基于知识元的学术论文创新性判断模型。其次，根据学术论文研究特点，构建理论与方法机器分类模型及知识元的抽取规则与抽取方法，建立规则库和知识语料库。最后，基于语义相似度计算方法，根据判断规则和相关权重对学术论文4 个维度的创新性进行评分。［结果/结论］基于知识元抽取的学术论文创新性评分系统的实证结果表明，该智能化评价方法具有一定的可行性，可为学术论文内容创新性智能化评价系统的最终实现提供方法借鉴。

Abstract: ［Purpose/significance］Innovation is the key factor of academic paper evaluation． Based on the knowledge element theory and machine learning theory and algorithm，this paper studies how to intelligently evaluate the innovation of academic papers from the content of paper． ［Method/process］Firstly，we constructed 4 knowledge element ontologies of academic papers including‘research problem ontology’，‘theory ontology’，‘method onology’and‘conclusion ontology’，and proposed the model of innovation evaluation． Secondly，we put forward the rules of knowledge element extraction． Word2vec and naive Bayes were used to classify the innovation of theories and methods of academic papers，and SVM model was used to build the rule base of knowledge element extraction． At last，on the basis of the construction of knowledge Meta base of academic papers，we proposed the basic methods of intelligent evaluation of research questions，theories，methods and conclusions of academic papers． We also constructed the process of intelligent evaluation of innovation of academic papers． ［Ｒesult/conclusion］The feasibility of the
methods is verified by the experiment and could provide the references for the realization of intelligently evaluation of academic paper．