Unmatched raw affiliation strings in OpenAlex (snapshot 2026-06-26, articles and reviews 2010–2026)

Raw affiliation strings that OpenAlex did not match to any institution, from the OpenAlex works snapshot of 26 June 2026 (works of type article or review with publication year 2010 or later). Ten tar archives (unmatched_strings_part01of10.tar to unmatched_strings_part10of10.tar; MD5 checksums in md5sums.txt) together contain 2,446 gzip-compressed tab-separated files, one per snapshot part, with no header and 39,455,458 rows: one row for each of the 23,313,935 authorships that have an affiliation string but no matched institution (state "unmatched"), and one row for each of the 16,141,523 authorships that have a matched institution and at least one further string that was not matched (state "matched"). Columns: work_id (OpenAlex numeric work id); author_id (OpenAlex numeric author id, 0 when absent); publication_year; country code of the authorship's matched institution, if any ("(none)" otherwise); state; the unmatched string(s), up to five per authorship, joined by " || ", each truncated at 300 characters, with tabs and line breaks replaced by spaces. Produced by code/02_raw_affiliation_scan.py in the companion record (https://doi.org/10.5281/zenodo.23851599). Derived from the CC0 OpenAlex snapshot. Intended for analysing which forms of affiliation string fail to match, by country, year and source.