ORCID Authorship Assignment: Define Acceptable Distance

[tobiasschweizer]

Currently, the assignment of an ORCID profile to a DOI (DOI is listed in an ORCIC profile) requires an exact match of first and last name of an author with the metadata provided by the DOI RA.

 # match ORCID profiles by name
    author_orcid = list(
        filter(lambda orcid: orcid.family_name == family_name and orcid.given_name == given_name, orcid_info))

To improve the chance of a match, an argument could be introduced defining the acceptable distance, i.e. what would still be considered a match also if the strings are not identical.

Choices:

• levensthein distance algorithm
• thefuzz, see https://gitlab.switch.ch/connectome/rdf2rdf/-/issues/65#note_180235

[tobiasschweizer]

Also integrate provenance information from ORCID profile (how the DOI was imported/assigned):

curl -L -H "Accept: application/orcid+json" "https://orcid.org/0000-0002-9016-1013" > orcid.json

."activities-summary".works.group[]."work-summary"[].source."source-name"

-> instead of requesting JSON-LD, request the ORCID native JSON format.

Get DOIs from native ORCID JSON: [."activities-summary".works.group[]."work-summary"[]."external-ids"."external-id"[] | select(."external-id-type" == "doi") | ."external-id-value"]

[tobiasschweizer]

eval cordis:

for doi in `jq -c '.[]' doi.json`; jq .$doi formatted.json >> eval.txt;