How Boundary-spanning Paper Sparkles Citation: From Citation Count to Citation Network

Previous studies diverged on whether boundary-spanning papers are cited more heavily, however, there is as yet no research that explains how boundary-spanning papers affect citation counts from a causal view and how they influence knowledge diffusion in citation network. To this end, we utilized Propensity Score Matching to clarify the relationship between the citation counts and the degree of boundary-spanning of 281,707 papers in the field of Astronomy (AS), Natural Language Processing (NLP), Library and Information Science (LIS) based on causal inference. We also adopted Sentence-BERT to compute the content similarity among seed papers, cited papers, citing papers, to imply how seed paper affects citation network when knowledge diffuses. According to the similarity between seed paper and citing paper, cited paper and citing paper, we defined paper's role in citation network as four types: Disseminator, Broker, Trigger, and Outlier. The major findings are as follows: (1) Papers with a higher degree of boundary-spanning are more likely to be cited heavily; (2) Disseminator and Outlier account for a larger proportion in three disciplines, while Broker and Trigger account for a smaller proportion; (3) The degree of boundary-spanning and citation counts of four types vary in three disciplines. This work, which reveals paper's value and role in citation network from the perspective of content, has implications to provide some enlightenment for the paper's evaluation.