ORCIDs mapped to PubMed authors

Dataset

Description

The dataset is based on a snapshot of PubMed taken in December 2018 (NLMs baseline 2018 plus updates throughout 2018), and for ORCIDs, primarily, the 2019 ORCID Public Data File https://orcid.org/.

Matching an ORCID to an individual author name on a PMID is a non-trivial process. Anyone can create an ORCID and claim to have contributed to any published work. Many records claim too many articles and most claim too few. Even though ORCID records are (most?) often populated by author name searches in popular bibliographic databases, there is no confirmation that the person's name is listed on the article. This dataset is the product of mapping ORCIDs to individual author names on PMIDs, even when the ORCID name does not match any author name on the PMID, and when there are multiple (good) candidate author names. The algorithm avoids assigning the ORCID to an article when there are no good candidates and when there are multiple equally good matches. For some ORCIDs that clearly claim too much, it triggers a very strict matching procedure (for ORCIDs that claim too much but the majority appear correct, e.g., 0000-0002-2788-5457), and sometimes deletes ORCIDs altogether when all (or nearly all) of its claimed PMIDs appear incorrect. When an individual clearly has multiple ORCIDs it deletes the least complete of them (e.g., 0000-0002-1651-2428 vs 0000-0001-6258-4628). It should be noted that the ORCIDs that claim to much are not necessarily due nefarious or trolling intentions, even though a few appear so. Certainly many are are due to laziness, such as claiming everything with a particular last name. Some cases appear to be due to test engineers (e.g., 0000-0001-7243-8157; 0000-0002-1595-6203), or librarians assisting faculty (e.g., ; 0000-0003-3289-5681), or group/laboratory IDs (0000-0003-4234-1746), or having contributed to an article in capacities other than authorship such as an Investigator, an Editor, or part of a Collective (e.g., 0000-0003-2125-4256 as part of the FlyBase Consortium on PMID 22127867), or as a "Reply To" in which case the identity of the article and authors might be conflated. The NLM has, in the past, limited the total number of authors indexed too. The dataset certainly has errors but I have taken great care to fix some glaring ones (individuals who claim to much), while still capturing authors who have published under multiple names and not explicitly listed them in their ORCID profile. The final dataset provides a "matchscore" that could be used for further clean-up.

Four files:
person.tsv: 7,194,692 rows, including header
1. orcid
2. lastname
3. firstname
4. creditname
5. othernames
6. otherids
7. emails

employment.tsv: 2,884,981 rows, including header
1. orcid
2. putcode
3. role
4. start-date
5. end-date
6. id
7. source
8. dept
9. name
10. city
11. region
12 country
13. affiliation

education.tsv: 3,202,253 rows, including header
1. orcid
2. putcode
3. role
4. start-date
5. end-date
6. id
7. source
8. dept
9. name
10. city
11. region
12 country
13. affiliation

pubmed2orcid.tsv: 13,133,065 rows, including header
1. PMID
2. au_order (author name position on the article)
3. orcid
4. matchscore (see below)
5. source: orcid (2019 ORCID Public Data File https://orcid.org/), pubmed (NLMs distributed XML files), or patci (an earlier version of ORCID with citations processed through the Patci tool)
12,037,375 from orcid; 1,06,5892 from PubMed XML; 29,797 from Patci

matchscore:
000: lastname, firstname and middle init match (e.g., Eric T MacKenzie vs
00: lastname, firstname match (e.g., Keith Ward)
0: lastname, firstname reversed match (e.g., Conde Santiago vs Santiago Conde)
1: lastname, first and middle init match (e.g., L. F. Panchenko)
11: lastname and partial firstname match (e.g., Mike Boland vs Michael Boland or Mel Ziman vs Melanie Ziman)
12: lastname and first init match
15: 3 part lastname and firstname match (David Grahame Hardie vs D Grahame Hardie)
2: lastname match and multipart firstname initial match Maria Dolores Suarez Ortega vs M. D. Suarez
22: partial lastname match and firstname match (e.g., Erika Friedmann vs Erika Friedman)
23: e.g., Antonio Garcia Garcia vs A G Garcia
25: Allan Downie vs J A Downie
26: Oliver Racz vs Oliver Bacz
27: Rita Ostrovskaya vs R U Ostrovskaia
29: Andrew Staehelin vs L A Staehlin
3: M Tronko vs N D Tron'ko
4: Sharon Dent (Also known as Sharon Y.R. Dent; Sharon Y Roth; Sharon Yoder) vs Sharon Yoder
45: Okulov Aleksei vs A B Okulov
48: Maria Del Rosario Garcia De Vicuna Pinedo vs R Garcia-Vicuna
49: Anatoliy Ivashchenko vs A Ivashenko
5 = lastname match only (weak match but sometimes captures alternative first name for better subsequent matches); e.g., Bill Hieb vs W F Hieb
6 = first name match only (weak match but sometimes captures alternative first name for better subsequent matches); e.g., Maria Borawska vs Maria Koscielak
7 = last or first name match on "other names"; e.g., Hromokovska Tetiana (Also known as Gromokovskaia, T. S., Громоковська Тетяна) vs T Gromokovskaia
77: Siva Subramanian vs Kolinjavadi N. Sivasubramanian
88 = no name in orcid but match caught by uniqueness of name across paper (at least 90% and 2 more than next most common name)

prefix:
C = ambiguity reduced (possibly eliminated) using city match (e.g., H Yang on PMID 24972200)
I = ambiguity eliminated by excluding investigators (ie.., one author and one or more investigators with that name)
T = ambiguity eliminated using PubMed pos (T for tie-breaker)
W = ambiguity resolved by authority2018
Date made availableMay 7 2021
PublisherUniversity of Illinois at Urbana-Champaign

Cite this