One sitting, four windows — from the Sinhala Tipiṭaka to Old Frisian
28 October 2026 · Nabu news
After a quiet week the library ran four windows in one sitting, and the shelves show it: 2,854,118 documents / 128,499,000 passages across 327 live sources (as of 7 November 2026).
The Pali canon arrived whole. The Sri Lankan tripitaka.lk edition — 563 documents, 456,967 passages of Pali with its Sinhala translation interleaved — joined the Theravada shelf, and the registry now records that its Pali is written in Sinhala script, so script-aware search and collation treat it honestly. Beside it the Mahāvyutpatti, the ninth-century translator’s dictionary, came in as a four-way bridge: 23,742 crosswalk edges joining Sanskrit, Tibetan, Chinese and Mongolian Buddhist vocabulary to the dictionaries already on the shelves.
The alignment hub grew by a Testament. The Bohairic Coptic Old Testament aligned against the hub’s other witnesses — 20 of 24 books, 10,493 verses, 64,188 new verse pairs, with the held-out books documented rather than forced — and the work uncovered a wiring-era defect: the Bohairic New Testament’s zero-padded chapter numbers had kept it out of co-alignment entirely. Folding the citation forms restored 32,894 pairs that should have been there all along, and the alignment index now rebuilds itself whenever its registry changes.
Old Frisian went live. The Corpus Oudfries — 177 witnesses of medieval Frisian law, 14,136 passages, 235,462 tokens with lemma and part of speech on every word — was reconciled exactly against the publisher’s own index. The Seventeen Statutes open the collection with the classic line Thit is thiv forme liodkest; manuscript dates and attributed composition eras ride separate timeline claims, and region, dialect and genre are all mined. It joins under the same research-private posture as the project’s other Dutch-institute corpora: readable at the desk, never redistributed.
Verify-first paid for itself twice. A suspected abbreviation misread in the South Indian inscription shelf turned out to be something better: the editors bracket damaged final digits (“Ś. 107[3]”), and the parser had been stopping at the bracket — reading Śaka 1073 as Śaka 107, a thousand-year error across exactly 48 documents, corroborated against the inscriptions’ own 60-year cycle names and now fixed with the lost digits read as honest decade ranges. And a hypothesis that trailing numerals in Egyptian dictionary spellings were undercounted repeat signs was refuted the same way — cross-examining two held corpora by token id showed they are numeral signs in their own right, so the sign tally stands.
Two more person authorities. The Digital Prosopography of the Roman Republic brought 4,876 persons of the Republic into the person index under its stated non-commercial license, joining the Chinese biographical database on the authority shelf; the Neo-Assyrian prosopography network (17,387 persons under CC BY-SA) is verified and queued next.
And the reading instruments kept widening. The Early New High German reference corpus opened its morphology to search: a census of all 190 held annotation files (2,061,753 tags) showed a closed vocabulary that now decodes into the same case/number/gender/tense names every other tagset uses — syncretic genders honestly searchable from both sides. Dead Sea Scrolls words now show the English gloss of their biblical-Hebrew lexeme, borrowed at read time through the editors’ own lexeme crosswalk (consonantal homographs kept apart). The Old Uyghur dictionary’s Manichaean, Christian and script marks became browsable facets; Chinese characters gained phonetic-series neighbours on the character card from the reconstructed Old Chinese shelf; and the Greek treebanks’ dialect labels — Attic, Ionic, Doric, Aeolic — now compose with their date bands, so a search can ask for Ionic prose of the classical period and mean both halves.