Open archives, journals and full texts

Besides the bibliographic index, cartolex can collect from an open archive (HAL) and a journal platform (SciELO), and improve the texts it found with open-access services. This page says what each brings, what it needs from you, and what it sends.

Who is found, and how

service	a person is found by	a name alone
HAL	their idHAL (a column of the people list)	proposes the author forms HAL has, with their idHAL when there is one
SciELO	their ORCID, when the journal printed it on the article	proposes the articles where the name appears

A name never adds a text by itself: a proposal waits for you to confirm it, the same way as an OpenAlex record: a HAL author form by its idHAL, a SciELO author by the ORCID the article shows (cartolex collect confirm my-project p000007 hal:ada-tavelin, or orcid:0000-…). The waiting list shows every service’s proposals side by side. Both services are read over the project’s window of years. SciELO cannot be searched by author: give it the journals to read (their ISSNs), or it reads a whole collection, up to 5,000 articles.

From HAL, each deposit brings its titles and abstracts in every language it has, its DOI, its type and year, and the structures (teams, labs, institutions, with the ones above them) its authors stated. From SciELO, each article brings its titles and abstracts in every language (often Portuguese, Spanish, English, French); the keywords its authors chose are not read, since the keywords come from the texts themselves.

One work, one text

The same work often comes from several services. cartolex joins them into one text when they share a DOI, a service’s identifier, or, for the same person, the same title with years at most one apart. When two services disagree (a year, a title), the value of the more reliable one is kept and the other is listed; nothing is replaced silently. A preprint and the article it became stay two texts, linked: the map reads the article only, so a work counts once. Every join is listed, with the rule that made it, in sources/merges.json. A work the services list twice under different DOIs (an article and its conference version, the same article twice) stays two texts in the tables, but the map reads it once: the Texts tab says how many duplicate texts are counted once.

Better texts

Missing abstracts are looked for, in this order, in SciELO, HAL, Europe PMC, bioRxiv and medRxiv, arXiv and OpenAlex.

Full texts are fetched only when you ask for them: first a structured text (the JATS of Europe PMC, bioRxiv, medRxiv and SciELO, the LaTeX source of arXiv), then a PDF (HAL’s file, or the open-access copy OpenAlex knows). A file that cannot be read is skipped on its own. By default the keywords are built from titles and abstracts; whether full texts count, and how much, is a setting of the build.

Full texts stay on your computer: they are never part of a shared map, a site, an export or anything cartolex sends.

What is sent, to whom

service	receives	pace
HAL	idHALs, names (for proposals), HAL identifiers, DOIs	2 requests a second
SciELO	journal ISSNs, article identifiers	1 a second
Europe PMC	DOIs, then Europe PMC identifiers	1 a second
bioRxiv, medRxiv	DOIs of preprints	1 a second
arXiv	arXiv identifiers	1 every 3 seconds, as arXiv asks
OpenAlex, open-access hosts	OpenAlex identifiers or DOIs, then the address of the open copy	as for OpenAlex; 1 a second per host

No text leaves the computer: only identifiers, DOIs, journal ISSNs and, for proposals, people’s names. The contact address you give goes with each request, as the services ask. Every host contacted and the kinds of data sent are recorded for the privacy summary.