Research On The Value Evaluation Of Academic Papers Based On Citation Context Analysis

Academic paper is an important carrier of scientific research achievements,and the value of academic paper is an important basis to measure the academic level and scientific research ability of researchers.However,the current evaluation method of academic papers mainly takes whether the papers appear in the reference list of other papers as the standard to measure the level of papers.All citations are regarded as equally important without distinction,and are uniformly expressed by the number 1.Accordingly,the total citations are used to represent the influence of academic papers to the maximum extent.In fact,the evaluation method based on the cited frequency can only show the number of times that the cited literatures are found and cited,but can not reflect the effect of the cited literatures on the cited literatures.Relative to the cited frequency,the citation content generated by citation behavior contains more profound information,such as citation motivation,citation emotion,citation purpose,etc.,these information for us to understand cited literatures ' contribution to the citing literatures,and on this basis to better assess the value of the cited literature provides an important auxiliary support.On the basis of examining the citation content,this dissertation first verifies the role of citation content features in distinguishing the importance of citation,and extracts the key features that can effectively distinguish the important citation,and on this basis,realizes the evaluation of the value of academic papers.(1)This dissertation proposes a classifier framework for distinguishing important citations from non-important citations,discusses the value of the content features in distinguishing important citations,and extracts the key feature factors for identifying important citations.The distinction of important citations helps to understand the value of the cited literatures to the citing literatures,that is,whether the cited literature are important citations relative to the citing literatures.Previous studies have explored the role of bibliometric characteristics of various citations in distinguishing important citations,but few studies have involved the characteristics of citation content.In this dissertation,we use natural language processing(NLP)technology to extract the semantic content features,and construct the feature space to distinguish important references together with the syntactic features.Three feature selection algorithms,PCC,Relief-F and entropy weight method(EWM),were introduced to sort the features.Three classifier algorithms,SVM,KNN and Random Forest,are used to determine the best feature subset.The experimental results show that both the syntactic and semantic features extracted from the citation content play a key role in distinguishing important citations,that is,these features are helpful in distinguishing the value of the cited literature to the citing literature.(2)The dissertation proposes an evaluation method of the value of academic papers based on the citation content feature.The value of academic papers mainly depends on its contribution in the scientific community,that is,the value of the academic papers to all citation literatures.Combined with citation content features extracted from the important citation experiment from the two levels of syntactic and semantics,this dissertation evaluates the value of academic papers from the perspective of citation,and compares it with the traditional evaluation results which only investigate the quantitative characteristics of citation frequency.How to distinguish the effect of different citation locations,different citation sentiments and different citation intentions on citation value evaluation is the key to reasonable quantitative eigenvalue and effective implementation of evaluation.In this dissertation,multiple linear regression method is introduced to calculate the weight of different citation locations,different citation emotion and different citation intention,and the comprehensive citation location scores,comprehensive citation emotion scores and comprehensive citation intention scores of the cited literature to the citing literature are obtained.At the same time,the deep learning method was introduced in this paper to calculate the subject similarity between cited literatures and citing literatures.In order to obtain the final evaluation value of the paper by integrating the indexes,this dissertation introduces the entropy weight method(EWM)to give weight to the evaluation indexes and realize the quantitative evaluation of the value of the academic paper.The results show that academic papers with the same or similar citation frequency have completely different citation value.This shows that only relying on a single quantitative index of citation frequency is not enough to completely distinguish the contribution of academic papers in the scientific community,and investigating the citation content of papers in the cited literature provides an effective way to evaluate the value of academic papers.This dissertation points out that the citation content is an effective data source to distinguish the importance of the cited literature to the citing literature,and to evaluate the value of the cited literature,which provides an important basis for the realization of a more reasonable value measurement of academic papers.