Release notes

A parser had split, cut or run together some affiliations: the tail of a parenthesis ("UK)"), degrees only, an email token, another editor's text or a whole board, a title remnant, an ORCID iD or a telephone number. Each now takes the value its board page gives, from the April 2026 cache or a Wayback Machine capture. Every correction passed a strict whitelist and was read against the page text; a second reading of 300 kept corrections drawn at random found no error.

6,324 positions take the affiliation their board page gives; 374 are left with none.

A name cleaner had cut "Dr" and "Prof" from the start of any word and dropped surnames that spell a credential ("Ma", "Peng"), and the scraper had cut name words that look like a credential and the last word after an initial. Each name is restored from the scrape output or the page text. Rows that held no person (a page label, an address, a journal title) or listed a person a second time, with the name glued to the city, are dropped.

5,814 positions change editor name and 52 ORCID iDs printed in an affiliation fill an empty orcid (orcid_source scraped_affiliation, a new value). 994 rows are dropped (920,966 remain), and with them 19 journals whose only row was a page label.

Headings such as "Assistant to the Editor" or "Editor-in-Chief's Assistant" had made their positions editors. They are now staff (role_std other), like the editorial office's assistants.

89 positions in 33 journals move from editor to other.

Every organization match of v6.1.0 was checked against ROR's data dump, and the wrong ones (another campus, another country, a namesake) are corrected. ROR's matcher had left many plain names blank ("Purdue University, USA"): every affiliation without an organization was read, or matched by a name matcher over the same dump that refuses whenever a namesake, a place or a generic name leaves a doubt.

70,657 positions change ror_id: 59,519 gain an organization, 9,915 move to another and 1,223 lose their match. 805,688 positions carry an organization, 93.2% of those with an affiliation (86.4% in v6.1.0), on 27,377 ROR records. About 0.66% of matched positions are estimated to remain wrong (95% interval 0.26% to 1.25%).

Gender is recomputed where the name or the country changed, or the stored value was stale. editor_key follows the name and the organization, so an editor whose blank rows now carry the organization of their other rows counts once.

2,902 positions change gender. 722,124 unique editors (731,244 in v6.1.0).

Every new organization match of the v6.0.0 ROR refresh was audited. The matches that share fewer than half of their organization's name words were all read by hand; of the others, those flagged by screens (a city or country that disagrees with the record, namesakes, generic names) were read and a sample was drawn. Most errors came from a ROR alias that is a common phrase ("The Council", "Chinese Academy" for Academia Sinica) or from a namesake (UT Southwestern read as Southwestern Medical Center in Lawton). Curated rules now give these affiliations the organization they name, clear matches to strings that name no organization in ROR (a place, a degree), and a watchlist of 47 organizations accepts a match only when the affiliation names them.

13.4% of the low-overlap positions were wrong, and 2.5% of the high-overlap positions were found wrong among those screened or sampled; about 0.3% more are estimated to remain wrong (95% interval 0.07% to 0.76%). 14,359 positions change ror_id: 13,433 move to another organization, 777 lose their match and 149 gain one. 747,516 positions (81.1%) carry an organization, on 24,966 ROR records.

An affiliation such as "Institute of Soil Science, Chinese Academy of Sciences, Nanjing" had been matched to the parent organization. It now carries the institute's own ROR record when the string names it in full and gives the institute's own city; system records are likewise sharpened to the campus or member a string names (University at Buffalo for the State University of New York). Institutes written at another site than their record's city, peers and second affiliations stay where they were.

The Chinese Academy of Sciences keeps 1,688 positions on its own record (5,245 in v6.0.0); its institutes carry the rest, and the University of Michigan now has the most positions on its own record.

Positions on an inactive or withdrawn ROR record move to the one active record ROR names as its successor (Tokyo Institute of Technology to Institute of Science Tokyo). ror_name is ROR's current name, and ror_country follows the ROR record, with its coordinates where the country changes (ICARDA, now in Lebanon). NIT Hamirpur keeps India: its ROR record wrongly says Bangladesh, and the error is reported to ROR.

408 positions on 36 ROR records move to their successor; 104 on 22 ROR records stay, as ROR names no successor or several. 6,420 positions of 300 renamed organizations take the new name; 8 positions change country.

ror_parent_ids and ror_parent_names give the organizations a position's ROR record names as its parents; ror_ancestor_ids and ror_ancestor_names give every organization above it, nearest first. A French joint unit typically has several parents (CNRS, a university, INSERM or INRAE). On this site an institution's page now counts the editors of its sub-units and breaks its total down by sub-unit, and the explorer finds an editor through any organization above their record. An editor of a joint unit counts under each of its parents, so institution totals do not add up.

157,933 positions, on 7,207 ROR records, have a parent in ROR. 87 columns in all (83 in v6.0.0).

The rule that standardizes roles looked for "editor-in-chief" and so missed plural headings such as "Editors-in-Chief": thousands of chief editors were labelled plain editors in every release since v2.7.0. Plural, spaced, abbreviated and French, German, Spanish and Portuguese chief headings now give editor_in_chief; a ranked chief (associate, deputy, assistant) gives deputy_editor and a section chief section_editor. Frontiers' Specialty Chief Editors lead one section and are section editors, except in single-specialty journals without a Field Chief Editor. Advisory roles and committees (MDPI's Topical Advisory Panel, consulting and advisory editors) are board members; academic, handling and action editors, who handle manuscripts, are associate editors; book review editors are section editors; editorial staff (managing editors, editorial assistants, social media editors) are other, except in journals where they are the only editor-type role.

role_std changes on 105,765 positions. Chief editors go from 17,239 to 19,409, and 1,791 journals gain their first; plain editor goes from 108,425 to 50,290 and other from 55,810 to 22,100; board members go from 470,656 to 533,679 and associate editors from 182,365 to 203,602.

Boards often list former, honorary, emeritus or founding editors beside the serving ones. role_is_former is true when the heading names such a role, in English, French, German, Spanish, Portuguese or Italian, so they can be set aside. Titles that people hold while serving (Professor Emeritus, Past President of a society sitting ex officio) are not flagged.

10,533 positions flagged, 1,913 of them chief editors.

board_eic_count gives the number of distinct current chief editors of each journal, former ones excluded. Where chief editors are at least half of a board of ten or more, the parser most likely carried a chief heading over the whole board; board_roles_suspect marks those journals and their count is left empty.

56 journals marked suspect. Of the others, 4,444 have no chief editor, 7,912 one, 1,896 two, 667 three or four and 218 five or more. These counts are by publisher and journal, so a title listed by two publishers counts twice.

The affiliation lookup had kept failed requests as "no organization" since v3.0.0. Every affiliation without an organization was asked again. An organization already in the dataset keeps its recorded name, country and coordinates; ror_city and ror_state come from ROR, and matches to an organization the affiliation does not name are cleared, as before, for the nine organizations on the watchlist.

120,494 positions gain an organization: 81.1% of positions carry one (68.1% in v5.0.0). 3,345 organizations are new to the dataset, and countries go from 189 to 194. 41 false matches to a watched organization are cleared, and 32 overrides correct 202 positions where ROR's matcher was wrong (Purdue University had Hue University of Education).

Gender inference uses the institution's country, so it is recomputed where the country changed, and where an earlier repair had left it stale. The editor identity key includes the organization, so editors listed at one institution under several affiliation texts are now one editor.

5,513 positions change gender (3,632 from unknown to a gender); women are 32.1% of resolved positions (32.2%). Unique editors go from 744,797 to 731,230, and editors on several boards from 111,980 to 119,210.

issn_print, issn_electronic and issn_type_source give each journal's print ISSN and E-ISSN as the NLM Catalog, DOAJ or Crossref types them, from records that list the journal's ISSN-L; each journal's page shows them. 45 more journals take their audited ISSN-L, among them Molecules (1420-3049), and one is corrected again (Eurasian Economic Review), with the metrics, indexing and field that follow.

Print or electronic ISSN for 14,263 journals: 13,865 electronic, 11,379 print, 10,981 both. Audited ISSN-L corrections now apply to 240 journals (195 in v5.0.0).

The enrichment matched each journal to an OpenAlex source by searching its title and taking the first result from the same publisher. Where a short title is part of a longer one, that result was often the longer journal: Breast Cancer Research took Breast Cancer Research and Treatment, and Development took Environment, Development and Sustainability. The source brings the ISSN-L, the journal metrics, open-access status, funders, indexing, NPI and field, so all of them were another journal's. An audit compared every journal's title and publisher with its source's, and every mismatch was resolved, by hand where the rules could not decide.

195 journals (8,833 positions). 153 take their own journal's source and everything that comes from it. The 42 for which no source can be trusted lose these columns, like the journals that never had a source. 93 of the 195 change field. Quartiles are ranked within each field, so they move for 256 journals: 144 of the corrected ones and 112 others. ISSN-Ls shared by two or more titles fall from 117 to 23, each the same journal listed twice.

In April 2026, after the indexing stage had run, a restore put back the ISSN-L, DOAJ status and NPI columns of many journals without recomputing their flags. Every journal's Web of Science and NPI flags are now recomputed by the enrichment's own rules from its stored ISSN-L, and its DOAJ flag from its stored DOAJ status. PubMed was checked again in the NLM Catalog for the journals whose ISSN-L changed after their first check.

Outside the corrected journals: indexed_wos turns true on 392 journals; indexed_doaj on 399 (false on 1); indexed_npi on 972 (false on 2); indexed_pubmed on 151 (false on 33) of the 3,129 rechecked. In all, 1,358 journals change an indexing column.

Mir@bel, which links its journal records to these boards, sent back corrected board URLs: SAGE's issue lists become its editorial-board pages, APA and Brill URLs point at the board itself, Lippincott journals move to Ovid, and MDPI's API addresses become its public pages. Each was accepted only as the old URL rewritten by the publisher's rule. The same rules were then applied to the publishers' other journals, after a sample of pages had been checked without crawling.

2,217 journals take a corrected source_url on 207,833 positions, 718 of them from Mir@bel's list. 14 corrections are held back: two Ovid pages that return 404, two APA digests with no board page, two journal-package listings, seven Brill journals whose board no capture shows, and the removed listing below.

SAGE's journal listing names two journal packages beside their member journals, and the scraper took them for journals; their boards repeat their member journals' boards. Economic and Industrial Democracy held the board of Ergonomics in Design, which SAGE's listing served at its address when the boards were collected, and none of its people is on Economic and Industrial Democracy's own board. The rows are removed. The journal will return with its own board at the next collection; until then, its page keeps its address and says why it is empty.

137 positions removed: 70 for Economic and Industrial Democracy, 67 for the two packages. Unique editors: 744,797 (v4.0.1: 744,842). 92 editors lose a board, and 76 of them are now on one board only.

v4.0.0, like v2.7.0 and v3.0.0 before it, left ror_city and ror_state empty on records whose institution was matched after the one-off city backfill of April 2026: mostly IEEE editors added in v2.3.0. The release build now takes both from the ROR record of every row's ror_id and refuses a bundle where they are missing or differ from ROR. No other value changes, and every count of v4.0.0 stands; data_version is 2026.4.

5,537 records gain a city and 5,405 a state (4,919 editors); the other 132, in Singapore and Hong Kong, stay without a state because their institutions' ROR records give none. Most affected: Beijing (360 records), Shanghai (141), Nanjing (110), Singapore (105). Re-run only city- or state-level analyses.

v3.0.0 took each journal's field from its most frequent OpenAlex topic. For large general journals that topic is an artefact of OpenAlex's classifier, which files letters, news and obituaries under catch-all topics: The Lancet was classed as Engineering, and Science, JAMA and The BMJ as Economics. The field is now the one held by most of the journal's research articles, used when it covers at least 40% of them or when the old topic was a catch-all; a hand check found thinner majorities unreliable, so those 1,450 journals keep their v3.0.0 field. Journals the Norwegian Register for Scientific Journals files as interdisciplinary are Multidisciplinary.

2,576 journals (183,173 positions, 19.9%) change field, counting each publisher's copy of a title separately. The Lancet, JAMA, NEJM and The BMJ are Medicine; Science, Nature, PLOS ONE and Scientific Reports are among 151 Multidisciplinary journals. 537 journals that had no field now have one.

The codebook has called this a per-field quartile since v2.0.0, but the published files carried one split across all journals: Q1 started at a two-year citedness of 2.75 in every field, so 42% of Engineering journals but 6% of Mathematics journals were Q1. Each field now has its own quartiles, computed from the journal's two-year mean citedness within its (corrected) field; journals with equal citedness share a quartile.

5,860 journals change quartile (339,930 positions). Q1 now starts at 1.29 in Mathematics, 2.25 in Medicine and 3.70 in Engineering.

v2.0.0 removed false ROR matches to seven organisations, most of them University of California editors matched to California Coast University. A later re-enrichment brought them back, and v3.0.0 carried 3,483 of them. The repair is re-applied: University of California rows move to the campus their affiliation names, and the other false matches are cleared. The seven are real institutions, so a match whose affiliation names the organisation is kept; the release build fails only on matches whose affiliation does not.

2,516 rows move to their UC campus (UCLA 477, Davis 395, Berkeley 393, …), 967 lose a wrong match, and 1 genuine match stays. ROR coverage moves from 68.19% to 68.08% of records.

OpenAlex had retired the source identifier recorded for some journals. Those that resolve through their ISSN-L now carry the current source, and their four journal metrics (two-year citedness, h-index, works and citations) come from it, which also feeds their quartile.

152 journals (6,463 positions).

editor_key includes the institution, so it changes on the repaired rows, and University of California rows now join the same people's existing records at their campus. Board institution and country statistics, and the multi-board columns, are recomputed for the affected journals and editors. Nothing else changes: the build fails if any column outside a fixed allowed set differs from v3.0.0, and only the columns named in these notes did.

Unique editors: 744,842 (v3.0.0: 745,125). Multi-board share: 15.04% (15.02%).