Where the data comes from
Theses come from university repositories, national registers and the shared repositories of many universities; who supervised whom comes from their records and from Wikidata. Papers and coauthors are asked of OpenAlex, and a person's own list of works of ORCID, as a page opens. Every source is read openly, under the site's own name, at the pace its host asks.
The kinds of source
The record is built from sources that publish what they hold openly, and from nothing else. They are of six kinds. Each gives something different, and each link a page draws says which kind it rests on, as where every link comes from explains.
- University repositories
- A university's own repository holds its theses as records: the title, the author, the year, the degree, and often the supervisor and the committee. Most run on a handful of platforms: DSpace, bepress's Digital Commons, EPrints, Pure, OPUS, and others a library built for itself. They are the record's main source.
- National registers
- Some countries keep one list of the theses their universities award. The record reads Portugal's RENATES, the register of the theses defended at Portuguese universities, and NUŠL, the Czech national repository of grey literature, which holds the theses of many Czech schools.
- Shared repositories
- A repository run for many universities at once: DiVA in Sweden, HAL in France, Dabar in Croatia, OhioLINK's ETD Center in Ohio, TDX in Catalonia, White Rose eTheses for three universities in the north of England. A thesis is counted under the shared repository that served it, with the school its record names.
- Wikidata
- The open database behind Wikipedia, which anyone may edit. Its statements of who was whose doctoral advisor are copied into the record whole and renewed regularly; a person's teachers and the identifiers other registers know them by are read from it as their card opens. A link Wikidata states is marked as stated and kept apart from the links a thesis record gives.
- OpenAlex
- An open index of research papers and their authors. A person's papers, coauthors and the authors they cite most are asked of it as their card opens, under an identifier that ties the person to it. The record keeps none of them and counts none, and a coauthorship is drawn beside the lineage, never as part of it.
- ORCID
- The register of researchers' own identifiers. An ORCID iD in a thesis record helps tell one person from another, and the works a person's ORCID record lists are shown on their page as the person's own claim, asked of ORCID only when a reader opens that part, never counted with the record's.
Beside these, for a physician with papers and no thesis, the public list of the clinicians enrolled in Medicare gives a degree, a school and a year, which their card shows as unconfirmed and which nothing counts.
How a repository is read
Most repositories share one protocol, the Open Archives Initiative's OAI-PMH, made so that a program can copy a repository's records without a person at the keyboard. A program asks the repository's protocol address for its records in a format it names, changed since a given day; the repository answers a page of records at a time, each page ending in a token that asks for the next. Every record carries an identifier that does not change and the day it last changed, so a later reading asks only for what is new.
Every repository that speaks the protocol serves simple Dublin Core, fifteen plain fields in which a supervisor, if named at all, is one contributor among others. Many serve richer formats beside it: ETD-MS, made for theses; MODS and MARC, from the library world; DSpace's own list of fields; xMetaDissPlus in Germany; UKETD_DC in Britain. The record asks each repository for the format that best says who supervised, and reads it as from a repository record to a link describes.
Each night the record asks every source it holds for what changed since it last asked, unless the host asked for longer between readings. Every record it is given is kept in Amazon S3, so a reading can be made again from the same records when a reader of them improves.
Not every source speaks OAI-PMH. Portugal's RENATES is read through its own search's list export, a school and a kind of degree at a time. Other national registers serve their lists through a web API (France's theses.fr), through SRU, the search protocol of library catalogues (the German National Library), or as files to download whole (Brazil's CAPES catalogue). Readers for these three are written; each enters the record only after its first full reading and a check of the supervisors it names against a source that already names them.
Where the supervisions are
The coverage page counts every doctoral supervision a source's own record states, by country and by the decade the thesis was written; the shelves list the sources behind them. The two charts below draw the newest count. A country's bar is as long as the record's reach into it, not as long as its universities' output: a country with few supervisions here may have many the record does not yet read.
Counted 2026-10-22. The full table is on the coverage page.
Counted 2026-10-22. The full table is on the coverage page.
The terms of reading
Every source is read openly and as itself. Each request names the project and its contact address in its user agent, the name a program gives a server, so a host's staff can see who is asking and write. Between two requests to one host the record waits the longest of three: the Crawl-delay the host's robots.txt asks, the pause the host asked for in writing, or one second.
A host that asks for a slower pace or fewer readings is given it, and what it asked and what it was told are kept with the source. The University of California's eScholarship, for one, is asked thirty seconds apart and read again every thirty days, from the one machine it was told the requests come from. Where a host was told that, no other machine asks it.
A robots.txt that disallows a path is not taken as a refusal: the record reads what an open protocol serves, at the pace robots.txt asks. A server that refuses, by answering 401, 403, 418 or 429, or with a page that challenges a browser to prove a person is there, is recorded as refusing access and is never asked another way: the record does not borrow another program's name, solve a challenge or change the address it asks from. A single file withheld among files that are served, as a thesis under embargo is, is passed over and the reading goes on; several in a row are a refusal.
Open aggregators, which gather the records of many repositories under terms of their own, such as OpenAIRE, BASE and CORE, are sources in their own right. A repository that refuses or cannot be reached may be read through one, at the aggregator's terms and pace, and its records are then counted as the aggregator's, never the repository's.
Finding repositories
No list names every repository that holds theses, so the record keeps a register of every source it has found, whatever its state: read and in the record, ready to be read, found and not yet reviewed, refusing access, offering web pages and no data a program can read, serving records that name no supervisor, or gone. The first page lists them by state under "Where the record stops", so a reader can see which sources are missing and why, and the shelves list the sources the record reads, country by country.
Repositories are found a country at a time. OpenAIRE, the European infrastructure for open science, lists the institutional repositories it knows by country; the countries with the most repositories the register lacks are taken first, and each repository's protocol address is found, asked to identify itself, and put in the register as found. A repository is asked for a sample of its records before it is read in full, to see whether they say who supervised at all.
To point the record to a repository it does not read, write to academicancestry@proton.me.