Anomalous Citations Detection In Academic Networks

Citation network analysis attracts increasing attention from disciplines of complex network analysis and science of science,where citations are usually used to assess the quality of a paper.How to objectively evaluate the citations with the continuous growth of publications is a source of great concern in related fields.However,there are some unreasonable citations in citation networks,i.e.,cited papers are not relevant to the citing paper.Existing research on citation analysis has primarily concentrated on the contents and ignored the complex relations between academic entities such as authors and journals.In this paper,we propose a novel research topic,that is,how to detect anomalous citations in citation networks.To be specific,we first define anomalous citations and propose a unified framework,namely ANCI,to detect anomalous citations in heterogeneous academic networks.ANCI is established based on matrix factorization and network representation,which considers not only the relevance of citation contents but also the relationships among academic entities including journals,papers,and authors.To verify the performance of ANCI,we construct two anomalous citation datasets based on Microsoft Academic Graph(MAG)and the Digital Bibliography & Library Project(DBLP).Then we compare the ANCI with other baseline methods and ANCI variants on datasets,respectively.Experimental results on these datasets demonstrate the effectiveness of the proposed method.For further study of anomalous citations,we make full use of citation context information in dataset CiteSeerX.First,we use citation context to classify each citation purpose in detail and calculate the relevance of citation context.Secondly,the difference between citing articles and cited articles is calculated.Combining the relevance of citation context and the difference of the article,we make the probability distribution statistics.Finally,the judgment of anomalous citations is made by setting the threshold.We conduct comparative experiments on embedding dimension selection,similarity measure and citation purpose classification respectively,and the experimental results confirm the rationality of the method.Detection of anomalous citations can distinguish the quality of citations,which not only help make a judgment on the impact of papers and scholars correctly but also carry profound significance for academic fairness.