ORCID-linked labeled data for evaluating author name disambiguation at scale

Abstract

How can we evaluate the performance of a disambiguation method implemented on big bibliographic data? This study suggests that the open researcher profile system, ORCID, can be used as an authority source to label name instances at scale. This study demonstrates the potential by evaluating the disambiguation performances of Author-ity2009 (which algorithmically disambiguates author names in MEDLINE) using 3 million name instances that are automatically labeled through linkage to 5 million ORCID researcher profiles. Results show that although ORCID-linked labeled data do not effectively represent the population of name instances in Author-ity2009, they do effectively capture the ‘high precision over high recall’ performances of Author-ity2009. In addition, ORCID-linked labeled data can provide nuanced details about the Author-ity2009’s performance when name instances are evaluated within and across ethnicity categories. As ORCID continues to be expanded to include more researchers, labeled data via ORCID-linkage can be improved in representing the population of a whole disambiguated data and updated on a regular basis. This can benefit author name disambiguation researchers and practitioners who need large-scale labeled data but lack resources for manual labeling or access to other authority sources for linkage-based labeling. The ORCID-linked labeled data for Author-ity2009 are publicly available for validation and reuse.

LAGOS‐AND: A large gold standard dataset for scholarly author name disambiguation

“…However, this facet has not received enough attention in previous datasets. To examine the gender distribution, we used Genni+Ethnea (Smith et al, 2013; Torvik & Agarwal, 2016), a widely used gender dataset containing 4,934,974 distinct names collected from PubMed (Kim & Owen‐Smith, 2021; Subramanian et al, 2021). We queried genders from Genni+Ethnea by author names to obtain the gender predictions of LAGOS‐AND and MAG.…”

“…Due to the limited name patterns, a small dataset will restrict the exploration of some data‐driven techniques. Note that, although some datasets such as GESIS‐DBLP, 6 SCAD‐zbMATH (Müller et al, 2017), and Kim‐PubMed (Kim & Owen‐Smith, 2021) have decent numbers of instances, they are limited in scopes (covered domains). For example, SCAD‐zbMATH is designed specifically for a mathematical domain database, zbMATH 7 .…”

Ethnicity‐based name partitioning for author name disambiguation using supervised machine learning

“…This dataset was generated by the researchers at the University of Michigan Institute for Research on Innovation & Science (UM‐IRIS) through matching selected name instances in publication records to an authority database, ORCID (Kim & Owen‐Smith, 2021 ). First, author full names (e.g., “Brown, Michael”) that appear 50 times or more in MEDLINE‐indexed publications published between 2000 and 2019 were listed.…”

A dangerous triangularization of conflicting values in academic publishing: ORCID, fake authors, and risks with the lack of criminalization of the creators of fake elements

“…Finally, editors who mandate ORCID from submitting, corresponding, or all authors as a prerequisite for submission to their journal need to be conscientious of the fact that this "integrity" tool for author identification and authentication is imperfect. Even though Kim and Owen-Smith (2021) heaped praise on ORCID, it is unclear how their large-scale meta-analysis of the ORCID database was unable to detect the accounts of fake authors, fraudulent authors, or false positives (i.e., identities that claim to be authors, but which are something or someone else, e.g., 35 cases reported in Teixeira da Silva (2021c). To fortify the concerns, a search for "fruit", including for specific fruits, yielded a fruitful set of results (Table 1).…”