Publication syncing

_data/publications.json is the merged list rendered in the home-page /#publications section. /publications/ redirects there. Its inputs are _data/orcid_publications.json (papers, preprints and conference works) and _data/scholar_publications.json (conference entries only). The profile URLs come from author.orcid and author.googlescholar in _config.yml. ORCID metadata takes precedence for matching entries; each merged work also records its sources.

Refreshing locally

python3 scripts/sync_orcid.py
python3 scripts/sync_scholar.py
python3 scripts/merge_publications.py

All three scripts use the Python standard library. Scholar fetches the public profile and reads additional citation pages for new conference entries when needed. It supports profile pagination. No API keys, browser automation, proxies or CAPTCHA solvers are used. Scholar can reject automated requests; an unavailable or malformed profile leaves the committed cache untouched. Unsuccessful detail requests leave an entry linked to its Scholar citation, with profile metadata available for duplicate matching.

You can also import a saved profile HTML file:

python3 scripts/sync_scholar.py --source /path/to/profile.html --no-details
python3 scripts/merge_publications.py

--details-dir optionally loads saved citation pages named by the final part of their citation_for_view ID, such as u-x6o8ySG0sC.html. This makes imports reproducible without more network requests.

Avoiding duplicates

The shared matching logic is in scripts/publication_metadata.py:

1. Compare normalised DOIs, removing resolver prefixes, case differences and URL encoding. Distinct DOIs remain separate versions.
2. If a DOI is missing, compare conference and abstract identifiers, including EGU/EMS codes embedded in DOIs and AGU abstract codes.
3. If identifiers are missing, compare normalised titles within the same conference and year. Punctuation, dashes and diacritics are ignored.

The same title at different meetings remains separate. Preprints remain separate from their published papers. Conference years are taken from abstract identifiers where available, so an EGU24 abstract deposited in 2025 is listed under 2024. DOI and abstract IDs are retained in the data and rendered as HTML metadata; abstract codes also appear beside conference titles.

Automatic refresh

The existing .github/workflows/sync_orcid.yml workflow now refreshes both sources daily at 06:23 UTC, tests the matching logic, merges the data, commits changes and explicitly requests a Pages rebuild. If one source is unavailable, the other can still update, using the last committed data for the failed source. Read the individual fetch steps in Actions to see source failures.

Checks: python3 -m unittest discover -s scripts -p 'test_publications.py'.

The Talks map reads the same merged conference records as Publications. New conference entries and title/link corrections appear on both pages after the next refresh/build. No separate title list needs updating. Author order is retained from Scholar when ORCID does not supply it. Both pages apply the shared _includes/conference-metadata.html annotations from _data/talkmap.yml. Use those annotations to confirm who presented, specify talk/poster format, and associate meetings with cities; these manual details survive source refreshes. See TALKMAP.md for examples.

References: ORCID identifiers, Google Scholar help.