A place ruled, the dictionaries grouped, and a canon with names
27 September 2026 · Nabu news
Three threads landed this week: the places program crossed from mining to ruling, the dictionary shelf learned its own family tree, and the Tibetan canon got its names.
The first ruled place. The library’s Chinese texts have been mined against the CHGIS historical gazetteer for a while — 2.97 million candidate sightings of nearly ten thousand distinct place-names across the Kanseki Repository’s five thousand texts. What was missing was judgment. The new review board lays each name out with its gazetteer candidates, dynastic-era discriminators and passage snippets, and verdicts now flow into a small curated registry: forty frequent-but-false names — bureaucratic titles, personal names, common nouns that happen to spell like villages — are stopped at the source, and 杜陵 — Du Ling, the Han-era county southeast of Chang’an where the classicist Tian He settled — became the first name ruled to a single gazetteer identity. Its 5,443 passage attestations are now first-class place links, readable from either end. One name down; the board holds the queue for the rest.
Dictionaries now group by lineage. The reference shelf has grown to 465 dictionaries and 2.6 million entries, shipped under dozens of language codes — Assyriologists’ period codes, Wiktionary’s etymology-only codes, plain ISO. A new precompiled census resolves every shelf through the lect registry, so a lookup in Latin now knows that the medieval-Latin shelf belongs to it, and the language card for Akkadian can show its whole dictionary family at once. The registry itself grew to meet the census: the full ORACC Akkadian dialect grid (Old Assyrian through Late Babylonian, Standard Babylonian as the literary register, the peripheral Akkadian of the Amarna age), Emesal — the Sumerian cult-lament register — and anchors for Ugaritic, Etruscan, Lepontic, Sabine, Canaano-Akkadian and Old Japanese, each with its classification and attestation recorded. Codes the registry cannot honestly place stay counted and labeled as unresolved rather than guessed.
The Tibetan canon has titles now. The Derge Kangyur and Tengyur arrived as exact woodblock reproductions — 4,560 texts identified only by their Tohoku catalog numbers. But the 84000 translation project’s files name every one of those texts in English and Tibetan, including the thousands not yet translated, whose placeholder records exist precisely to hold the titles. A crosswalk now reads those headers and titles 4,451 of the 4,560 documents: Toh 767 is no longer a bare number but the Tantra of the Supreme Dancer of the Yakṣas (gnod sbyin gar mkhan mchog gyi rgyud). And the same files’ back matter — glossaries the parser had always stepped around — yielded 56,794 term records linking Tibetan, Sanskrit and English (garuḍa / གསེར་འདབ / gser ‘dab), now a searchable shelf of its own beside the Mahāvyutpatti. The entries carry 84000’s CC BY-NC-ND grant, credited as always.