科研论文创新性与引文影响力相关性研究——基于语义视角的测度

Research on the Correlation between Creativity and Citation Impact of Scientific Research Paper:Measurement from Semantic Perspective

[目的/意义]设计了一种基于语义内容的论文创新性事前测度方法,揭示论文创新性与学术影响力之间的关系及其机理。[方法/过程]将单篇论文创新性分为新颖性和常规性两个特性;通过Word2Vec模型训练关键词向量,融合关键词语义计算论文新颖性及常规性;进一步结合统计回归分析方法分析了论文创新性对被引次数的影响。[结果/结论]以WoS收录的自然语言处理研究论文为对象进行分析,结果表明:(1)文章提出的论文新颖性和常规性的测度方法充分考虑了文本语义,在论文发表之初就可以获得其创新性评价结果,且创新性得分结果相对合理;(2)回归分析结果表明,科研论文越新颖其被引次数越少;(3)在新颖性相同的情况下,科研论文的常规性越高则被引次数越多,常规性对论文被引次数的影响程度相对于新颖性更强;(4)引文影响力更适用于高常规化科研领域的评价问题。

[Purpose/significance]This study aims to design a method which takes semantic content into account for scientific research creativity measurement,and enables the evaluation of creativity once the paper is settled down.Furthermore,this study intends to reveal the relationship and mechanism between the creativity of scientific research and academic influence.[Method/process]Firstly,the creativity of a paper is specifically explained into two characteristics:novelty and conventionality.Then,the word2vec model is used to obtain keyword vectors to represent the semantic information,which are used to calculate the scores of papers’novelty and conventionality.Furthermore,statistical regression analysis is employed to analyze correlation between the impact and creativity of the paper.[Result/conclusion]Analyzing research papers about natural language processing that are indexed in the Web of Science,it is found that the method of papers’novelty and conventionality measurement proposed in this study can reveal the text semantics and get reasonable scores.These scores effectively represent the scientific research creativity and can be obtained at the beginning of publication.Besides,the results of regression analysis indicate that the more novel a scientific research paper is,the fewer it is cited.When two scientific papers have the same novelty score,the higher conventionality,and the more citations.Compared with novelty,the conventionality has stronger impact on citations.Meanwhile,citations are more appropriate for the evaluation of mature scientific fields.