A boosted-trees method for name disambiguation

Jian Wang (), Kaspars Berzins (), Diana Hicks (), Julia Melkers (), Fang Xiao () and Diogo Pinheiro ()
Additional contact information
Jian Wang: Georgia Institute of Technology
Kaspars Berzins: Georgia Institute of Technology
Diana Hicks: Georgia Institute of Technology
Julia Melkers: Georgia Institute of Technology
Fang Xiao: Georgia Institute of Technology
Diogo Pinheiro: Georgia Institute of Technology

Scientometrics, 2012, vol. 93, issue 2, No 10, 411 pages

Abstract: Abstract This paper proposes a method for classifying true papers of a set of focal scientists and false papers of homonymous authors in bibliometric research processes. It directly addresses the issue of identifying papers that are not associated (“false”) with a given author. The proposed method has four steps: name and affiliation filtering, similarity score construction, author screening, and boosted trees classification. In this methodological paper we calculate error rates for our technique. Therefore, we needed to ascertain the correct attribution of each paper. To do this we constructed a small dataset of 4,253 papers allegedly belonging to a random sample of 100 authors. We apply the boosted trees algorithm to classify papers of authors with total false rate no higher than 30% (i.e. 3,862 papers of 91 authors). A one-run experiment achieves a testing misclassification error 0.55%, testing recall 99.84%, and testing precision 99.60%. A 50-run experiment shows that the median of testing classification error is 0.78% and mean 0.75%. Among the 90 authors in the testing set (one author only appeared in the training set), the algorithm successfully reduces the false rate to zero for 86 authors and misclassifies just one or two papers for each of the remaining four authors.

Keywords: Name disambiguation; Common names; Classification tree; Boosted trees (search for similar items in EconPapers)
Date: 2012
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (22)

Downloads: (external link)
http://link.springer.com/10.1007/s11192-012-0681-1 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:scient:v:93:y:2012:i:2:d:10.1007_s11192-012-0681-1

Ordering information: This journal article can be ordered from
http://www.springer.com/economics/journal/11192

DOI: 10.1007/s11192-012-0681-1

Access Statistics for this article

Scientometrics is currently edited by Wolfgang Glänzel

More articles in Scientometrics from Springer, Akadémiai Kiadó
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().