{
 "corpus": "pessoa",
 "seat": "EA-CORPORA-05/03",
 "origin": {
  "repo": "pt.wikisource.org",
  "method": "API list=categorymembers per heteronym category, then prop=revisions raw wikitext",
  "categories": [
   "Categoria:Alberto Caeiro",
   "Categoria:Ricardo Reis",
   "Categoria:Álvaro de Campos",
   "Categoria:Fernando Pessoa"
  ]
 },
 "edition": "Portuguese Wikisource transcriptions of the heteronymic corpus, taken by declared category",
 "license": "Public domain (author d. 1935); Wikisource transcription under CC BY-SA",
 "fetched": "2026-08-29",
 "works": {
  "count": 783,
  "by_voice": {
   "caeiro": 123,
   "reis": 60,
   "campos": 107,
   "pessoa-ortonimo": 493
  },
  "units": "one file per poem, prefixed by voice",
  "per-work": "text/<voice>--<Title>.txt",
  "originals": "original/ptwikisource-<voice>.tar.gz"
 },
 "normalization": "raw wikitext taken via prop=revisions; poem bodies extracted from <poem> tags; templates, category links and wiki markup stripped; line breaks preserved. The prop=extracts API returns EMPTY for the short lyrics because their text sits inside poem templates it does not render — the raw-wikitext route is required and is the reason an extracts-based fetch of this corpus silently yields almost nothing.",
 "verified_loci": [
  "O Guardador de Rebanhos (Caeiro)",
  "Odes, Livro Primeiro (Reis)",
  "Ode Marítima; Tabacaria (Campos)"
 ],
 "voice_partition": {
  "_note": "THE PARTITION IS THE SOURCE'S, NOT THE SEAT'S. Each poem is filed under the heteronym its Wikisource category declares. That partition is exactly the object under examination, so it is recorded as given and never adjusted to suit a result.",
  "caeiro": "123 texts, 16,123 words — the master the other voices acknowledge; the nearest thing in the modern cases to an orthonym",
  "reis": "60 texts, 5,548 words",
  "campos": "107 texts, 41,142 words",
  "pessoa-ortonimo": "493 texts, 48,636 words — the position bearing the civil name; under the archive's functional definition the orthonym is whichever position bears the founding gesture, which in this configuration is contested between Caeiro (whom the others acknowledge as master) and Pessoa-himself (who bears the name). The seat records both and settles neither."
 },
 "not_in_seat": "THE HOLD-OUT IS ANALYTICAL, NOT ARCHIVAL (corrected 2026-08-29). The first version of this seat withheld the orthonym's pages so that a leave-one-out test would remain possible. That reasoning was wrong and the correction is recorded rather than quietly applied. Withholding a voice at seat time does not protect the test — it prevents its validation. With three voices seated, holding one out leaves two, and two points cannot triangulate a third; worse, it yields one fold where the design needs four, and makes it impossible to check whether the procedure recovers a voice nobody considers special. If a method only ever succeeds on the position it was built to find, that is a story about that position and not an instrument. All four voices are therefore seated, and each is withheld in turn at analysis time. Three pages of the 496 listed under Categoria:Fernando Pessoa did not resolve and are absent; that shortfall is stated rather than closed over.",
 "why_seated": "Pessoa is the third term and the only one of the three with documented ground truth: the letter to Adolfo Casais Monteiro of 13 January 1935 attributes the heteronyms, so the grouping is known rather than inferred. That makes this corpus the positive control for any instrument claiming to detect designed differentiation, and nothing measured on Plato and Aristotle means anything until it has been calibrated here first.",
 "lines_total": 20486
}