The local structure of citation networks uncovers expert-selected milestone papers
作者：
Highlights：
• We compare the performance of global metrics and their local variants in the identification of expert-selected milestone papers on two distinct citation datasets.
• We obtain a family of local variants of PageRank (LeaderRank) with a tunable degree of locality and find they can generate highly correlated scores with the scores by the original metrics.
• The local variants of PageRank (LeaderRank) perform similarly well as the original global PageRank (LeaderRank) algorithms in the identification of seminal papers.
• The local variants of PageRank (LeaderRank) perform significantly better than local metrics such as citation count, h-index and semi-local centrality.
• Compared to network-based global metrics, the proposed local estimates provide a better trade-off between performance and computational efficiency.
摘要
•We compare the performance of global metrics and their local variants in the identification of expert-selected milestone papers on two distinct citation datasets.•We obtain a family of local variants of PageRank (LeaderRank) with a tunable degree of locality and find they can generate highly correlated scores with the scores by the original metrics.•The local variants of PageRank (LeaderRank) perform similarly well as the original global PageRank (LeaderRank) algorithms in the identification of seminal papers.•The local variants of PageRank (LeaderRank) perform significantly better than local metrics such as citation count, h-index and semi-local centrality.•Compared to network-based global metrics, the proposed local estimates provide a better trade-off between performance and computational efficiency.
论文关键词：Citation networks,Milestone papers,PageRank,Bibliometric indicators
论文评审过程：Received 1 March 2021, Revised 30 September 2021, Accepted 18 October 2021, Available online 3 November 2021, Version of Record 3 November 2021.
论文官网地址：https://doi.org/10.1016/j.joi.2021.101220
