About

How this site is built: publication metadata gathered from ORCID, CrossRef, OpenAlex, Semantic Scholar and doi2bib, clustered into topics with BERTopic and rendered with Quarto.

Most of this site is generated from my publication record rather than written by hand. The publication list, the per-paper pages and the research landscape are built from metadata fetched through public APIs, so they stay current without anyone editing a bibliography.

Where the metadata comes from

ORCID holds the authoritative list of works; everything else enriches it. Where two sources disagree, each field records which one it came from.

Source What it contributes
ORCID the list of works itself
CrossRef authors, affiliations, journal, volume and pages
OpenAlex open-access status and links, plus authors and abstracts where CrossRef has gaps
Semantic Scholar abstracts and TL;DR summaries
doi2bib BibTeX entries
Altmetric the attention badge on each paper page, fetched in your browser rather than at build time

Not every field comes from an API. Where no source has a usable abstract I write one myself, and a BibTeX entry is occasionally rendered from the metadata when doi2bib has none. The footer of each paper page names the origin of every field it shows, crediting manual entry or this site where that applies, so the provenance is visible per paper rather than asserted in the summary above.

How the research landscape is built

Abstracts are lemmatized with NLTK, embedded with the all-MiniLM-L6-v2 sentence-transformers model, projected onto two dimensions with UMAP and clustered into topics with BERTopic. Papers without a usable abstract fall back to their title. The result is the scatter plot on the research landscape page, and the topic label attached to each paper.

Tooling

The backend is written in Python with the my-scientific-profile library. The frontend uses Quarto. The publication list is one of Quarto’s own filterable listings; the landscape plot, and the tables on the landscape and topics pages, are Observable JS reading a CSV from this site and running in your browser, so none of that data goes anywhere else.

The source for this site lives at tbereau/tbereau.

Analytics and third parties

Visits are counted with GoatCounter, which sets no cookies and does not track visitors across sites.

Some requests do leave your browser, though, and a page that promises otherwise should say which. Each paper page loads its Altmetric badge from Altmetric when you open it, so they see that the page was viewed. A few pages also pull MathJax from a CDN, wherever Quarto judges that something needs mathematical typesetting. Everything else — fonts, styles, the site’s own scripts and the publication data — is served from here.