Research On The Improvement Of Position Weight In Co-citation Analysis

Co-citation analysis is a commonly analysis method in the field of intelligence,it can be used to reveal the clustering relationship between papers.With its simplicity and validity,it is closely concerned by the scientific metrology community in terms of papers clustering.In the past research of co-citation analysis,most people only use the co-citations frequency to measure papers' closeness,but they don't consider that the importance of different co-cited positions of co-cited papers is different,and the co-citations frequency of different co-cited positions of co-cited papers can be used to represent papers' closeness.From the perspective of paper's structure,the relative positions of the co-cited papers appearing in the citing papers can be divided into sentence level,paragraph level,section level,chapter level and part level.If the co-citations frequency of the co-cited papers is weighted according to their relative positions,not only the degree of closeness between the co-cited papers can be more accurately described,but also obtains more refined and more reasonable papers clustering results.When judging the weights of different co-cited positions,considering that the subjectivity of people judgment of most of the current researches will have an impact on the papers clustering results,this paper proposes that several experts use the analytic hierarchy process to judge the weights of different co-cited positions,then integrates the weights given by these experts to obtain an improved weights.The purpose of this action is to reduce the error caused by people subjective judgment when people using the analytic hierarchy process by using the group judgment method.At last,comparing with the papers clustering result obtained by using the previous weights to examine the scientific and the rationality of the papers clustering results obtained by the improved weights.