An open dataset of all the issues of this repository is available here : https://data.enseignementsup-recherche.gouv.fr/explore/dataset/openalex-affiliations-corrections/information/. This dataset is updated every day at 2 AM.
Exhibits cases where OpenAlex affiliation detection could be improved
In OpenAlex, an alignment pipeline is implemented to link raw harvested affiliations (particularly via web crawling I suppose) to standardised affiliations, described by a standard display_name and potentially an id, ror and country_code.
Each issue listed in this repo corresponds to ONE raw affiliation string whose RoR matches is inaccurate or missing.
October 2023 feedback
In mismatch_country_asof_20231017.jsonl, we list some raw_affiliation_string that we detected to provoke country mismatch (and then of course RoR mismatch). There are few (around 50 + variants) but they affect many publications. We provide also some 'contaminated' DOIs, all published since 2013.
A few examples:
-
all affiliations that contain a researcher ORCiD are linked to the US (assuming an affiliation to ORCID) ex: https://api.openalex.org/works?filter=doi:10.21203/rs.3.rs-480042/v1
-
University College London linked to Australia ex: https://api.openalex.org/works?filter=doi:10.1609/icwsm.v15i1.18053
-
Sorbonne University linked to UAE ex: https://api.openalex.org/works?filter=doi:10.21468/scipostphys.10.3.054
-
Dnipro State Medical University linked to France ex: https://api.openalex.org/works?filter=doi:10.26641/2307-5279.25.1.2021.231404
-
also very vague strings are matched, like "Independent Researcher" linked to Great Britain ex: https://api.openalex.org/works?filter=doi:10.1016/j.qref.2021.01.019
Some of the mismatches we detected seem explainable (like the bug for ORCiD), but others seem very weird, like the Dnipro State Medical University matched with a Valeo RoR ?
First feedback (now deprecated)
This repo lists a sample of cases where the country of affiliation present in OpenAlex appears potentially incorrect. We have used our own affiliation-matching tool to detect these cases, and this automatic tool itself is not perfect. Nevertheless, we believe that the vast majority of the cases raised here are of interest. We present this data with the following fields
-
raw_affiliation_string: raw affiliation string as present in OpenAlex
-
openalex_work_id: work id in OpenAlex in which the raw_affiliation_string has been found
-
openalex_country: country_code detected by OpenAlex for this raw_affiliation_string
-
openalex_display_name: display name of the institution detected by OpenAlex
-
matched_country: country_code matched by our affiliation-matching tool
We have separated the data into two files:
missing_country_code.old.csv: cases where OpenAlex does not provide a country_code
e.g "KU, Leuven, Leuven, Belgium" from https://openalex.org/W3085273257 has no country_code in OpenAlex (should be 'BE')
mismatch_country_code.old.csv: cases where OpenAlex MAY not provide an accurate country_code. Again these detections were done automatically and does contain errors for sure, it is NOT a golden dataset, but again, we believe that the vast majority of the cases raised here are of interest.
e.g "ANDRA, Ci2A, Soulaines-Dhuys, France" from https://openalex.org/W2802150657 is matched by OpenAlex to "Australian National Drag Racing Association", country_code AU whereas it should be matched to country 'FR'.