Citation Oriented AuthorRank for Scientific Publication Ranking
Jinsong Zhang; Xiaozhong Liu
It is now generally accepted that an article written by influential authors often deserves a higher ranking in information retrieval. However, it is a challenging task to determine an author&rsquo:s relative influence since information about the author is, much of the time, inaccessible. Actually, in scientific publications, the author is an important metadata item, which has been widely used in previous studies. In this paper, we bring an optimized AuthorRank, which is a topic-sensitive algorithm calculated by citation context, into citation analysis for testing whether and how topical AuthorRank can replace or enhance classical PageRank for publication ranking. For this purpose, we first propose a PageRank with Priors (PRP) algorithm to rank publications and authors. PRP is an optimized PageRank algorithm supervised by the Labeled Latent Dirichlet Allocation (Labeled-LDA) topic model with full-text information extraction. We then compared four methods of generating an AuthorRank score, looking, respectively, at the first author, the last author, the most famous author, and the &ldquo:average&rdquo: author (of a publication). Additionally, two combination methods (Linear and Cobb&ndash:Douglas) of AuthorRank and PRP were compared with several baselines. Finally, as shown in our evaluation results, the performance of AuthorRank combined with PRP is better (p <: 0.001) than other baselines for publication ranking.
Показать больше [+] Меньше [-]Библиографическая информация
Эту запись предоставил Multidisciplinary Digital Publishing Institute
