You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Needs further review and testing. OpenAlex supports two distinct search modes:
graph: searches a pre-computed citation graph based on the citation lists that have been gathered from papers. This is really inexpensive API wise and finds a lot of results
text: full text search, which can find string matches in data availability statements and methods, etc. This is far more expensive via the API, but produces different results from the graph search.
I ran a comparison of 200 randomly chosen DOIs from the ADC, and here's the results of what was found only in the text search, only in the graph search, and which were found by both:
Update after review of #46: the functional changes are below.
OpenAlex is opt-in in citation_search(). It spends credits, so it runs only when named in sources. Use citation_search_openalex() directly for cost control and resumable searches.
Resumed searches log each match once. Runs completed in installments no longer append duplicate records to the review logs.
Data relations are classified as a separate step. The search no longer calls DataCite. classify_data_relations(log_dir) can label dataset-to-dataset links afterwards as version, derivation, composition or citation. It doesn't affect citation results or counts; it's there for future work on data-to-data citations.
Stricter credit and rate-limit handling. Every billed response counts toward max_credits. Concurrent shards honour each other's rate-limit pauses during a run, and a recorded pause is never shortened.
Tests no longer spend credits by default. Live OpenAlex tests are skipped on CRAN and CI, and run locally only with SCYTHE_TEST_OPENALEX=true. Live DataCite tests are also skipped on CRAN and CI.
Activity
mbjones commentedon Oct 2, 2026
@jeanetteclark I started prototyping this tonight, and will try to push a branch tomorrow.
mbjones commentedon Oct 8, 2026
Did a pretty complete implementation in branch https://github.com/DataONEorg/scythe/tree/feature-44-open-alex
Needs further review and testing. OpenAlex supports two distinct search modes:
I ran a comparison of 200 randomly chosen DOIs from the ADC, and here's the results of what was found only in the text search, only in the graph search, and which were found by both:
See the vignette for openalex for a better explanation: https://github.com/DataONEorg/scythe/blob/feature-44-open-alex/vignettes/openalex-arctic-data-center.Rmd
mbjones commentedon Oct 10, 2026
Update after review of #46: the functional changes are below.
citation_search(). It spends credits, so it runs only when named insources. Usecitation_search_openalex()directly for cost control and resumable searches.classify_data_relations(log_dir)can label dataset-to-dataset links afterwards as version, derivation, composition or citation. It doesn't affect citation results or counts; it's there for future work on data-to-data citations.max_credits. Concurrent shards honour each other's rate-limit pauses during a run, and a recorded pause is never shortened.SCYTHE_TEST_OPENALEX=true. Live DataCite tests are also skipped on CRAN and CI.🤖 Generated with Claude Code
mbjones commentedon Oct 14, 2026
Closed and merged with PR #48