Skip to content

Why Voltaire is not in the registry: people enter through senses, not through credits — 6,674 persons are word-referents; decide on-demand minting, a people seed, or both #165

Closed as not planned

Description

@holden

The question

Voltaire is not in the registry. Neither is Kurt Vonnegut, Picasso, Shakespeare, Einstein or Darwin. Mark Twain, Oscar Wilde, Napoleon and Sigmund Freud are. The owner asked (2026-09-23, on #158) how that is possible when we have pulled entries for "basically everything", whether huge swaths are missing, and whether people should be created on demand. This issue answers the first two from the database and puts the third to a decision.

Refs #158 (quotes research; build 2 there mints creators on demand), #164 (creator identity, which mints creators on demand), #105 (exemplars name five people, none in the encyclopedia), #74 (the encyclopedia model), #73 (unified identities for people, authors, works).

What the database says (devils_dictionary_v2, 2026-09-23)

Entity kind Rows
concept 66,160
taxon 17,097
work 8,241
person 6,674
organization 1,523
place 617

Every one of the 6,674 persons has a Wikidata QID. Seven of them were produced by a discovery provider (six by the Wikidata absorb's own entity path, one by Artsy). The rest arrived through words: the current assertions whose object is a person are about 6,667 (a Wikipedia article about that person), lexeme_entity_candidate 4,001 (a word that is also a name, offered as a referent) and refers_to 956 (a sense that names the person). Bierce and Johnson are seeded by hand in Sources.Catalog.people/0.

So the people in the registry are the people who are the referent of a dictionary word: Napoleon (the pastry, the coin, the complex), Freud (Freudian), Twain, Wilde and 6,600 others such as Mavado, Muster and Una Merkel, who are there because a lemma matched an article title. Voltaire is not a word in any absorbed dictionary, so nothing ever asked for him. This is not a missed import. It is the #74 model doing what it was told: entities enter through senses, and no source we absorb is a source of people.

We are therefore not "missing huge swaths" of what we set out to hold. We are missing the entire class of people who are not referents — which is most authors, artists and speakers — because nothing yet credits them. The moment a source does (a Met artwork by Picasso, a quotation by Voltaire), the person is needed and absent.

The options

  1. On demand, from credits only (what Quotes research: which sources are reliable, measured on Voltaire and Vonnegut — Wikiquote re-tested through Parsoid, Wiktionary citations already absorbed, the aggregators fail corroboration, a keyless primary-text check (toward a :quote row) #158 build 2 designs). A person is minted when a provider credits them by QID and the registry lacks them: label, dates, description, person_details, the QID exclusive, through the Wikidata client at its pacing. Nobody enters without a reason to be on a page. Cost: one Wikidata call per new person, at read time of the first result that credits them, budgeted like any other call.
  2. A bounded people seed for the classes we will surely credit: every human with an English Wikiquote page (the theme pages and author pages Quotes research: which sources are reliable, measured on Voltaire and Vonnegut — Wikiquote re-tested through Parsoid, Wiktionary citations already absorbed, the aggregators fail corroboration, a keyless primary-text check (toward a :quote row) #158 will read), every Gutenberg author with a QID (P2034 on the work, P50 to the author), every Met and Open Library creator with a crosswalk. Tens of thousands at most (verify with the Wikidata client; the SPARQL count route answered 502 on 2026-09-22), one absorb run, a manifest under priv/, tier by era. Cost: a new absorb source that is a source of people, which the catalog and scorecard A1 do not yet know how to grade.
  3. Both: the seed for the classes we can name, on demand for the long tail. The seed makes the first page fast and the on-demand path keeps the model honest.

Recommendation: option 3, with option 1 first because build 2 needs it regardless, and the seed as a second issue once a Quotes shelf and an Artworks shelf can show what it is for.

Decisions for the owner

  1. Is a person who nothing yet credits allowed in the registry at all (option 2), or does every person need a reason on a page (option 1 only)?
  2. If seeded: which classes, in which order — Wikiquote authors, Gutenberg authors, museum constituents?
  3. Does a minted person get a page immediately (/entities/<id>/<slug> already renders any entity), or only once something links to them? Today an entity with no assertions renders empty.

Acceptance for this issue

Activity

  1. holden commented on Oct 21, 2026

    @holden
    MemberAuthor

    Folded into #217 (sense-level coverage), on the owner's word, 21 October 2026. The content stands; see the consolidation comment there.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions