[ronaldtse]

Summary

Batch 5 of the ALA-LC romanization encoding campaign (misc scripts): eight new systems transcribed in full from the loc.gov source documents.

System	File	Source	Rules	Embedded tests
Armenian (2022)	alalc-hye-Armn-Latn-2022.isc	armenian.pdf, 7 pp	134	12
Coptic (2014)	alalc-cop-Copt-Latn-2014.isc	coptic.pdf, 3 pp	103	16
Deseret (2016)	alalc-und-Dsrt-Latn-2016.isc	deseret.pdf, 2 pp	80	81
Cherokee (2012)	alalc-chr-Cher-Latn-2012.isc	cherokee.pdf, 1 p	85	18
Manchu (2012)	alalc-mnc-Mong-Latn-2012.isc	manchu.pdf, 5 pp	39	43
Moroccan Tamazight (2012)	alalc-zgh-Tfng-Latn-2012.isc	tamazight.pdf, 1 p	43	14
Tamashek (2013)	alalc-tmh-Tfng-Latn-2013.isc	tamashek.pdf, 1 p	40	26
Vai (2010)	alalc-vai-Vaii-Latn-2010.isc	vai.pdf, 3 pp	240	11

Every table row and note of each document is transcribed: letter tables with upper/lower case pairs, diacritic and digraph rows, punctuation transcription rules, and the Armenian and Coptic letter-numeral tables. Position-dependent values, closed exception lists, and other behaviors a character map cannot express are documented in each map's metadata notes rather than silently dropped.

Decisions and flags for review

• Deseret language id: the table scopes the Deseret script (no ISO 639 language), so the language field carries und. Naming follows the corpus authority-language-Script-Target-year shape; happy to re-id if a different convention is preferred.
• Armenian: primary renderings encoded; West Armenian bracketed variants documented. և → ev (per the table's own romanized examples); Եւ/Եվ → Ew/Ev. Note 2/Note 4 position-dependent values (ը→Y initially, յ→Ḥ stem-initially) and the Note 6 eʹv exception list are notes-only. ՞/՜ require sentence-final repositioning — not encodable, documented.
• Coptic: upper-case romanization cells for Ⲻ and Ⳃ are blank in the source (lower-case n̈/ç̈); the map supplies N̈/Ç̈ as mechanical case counterparts — flag. The numeral-6/6000 alternate glyph (absent from the text layer, substitute font) is identified visually as Greek stigma and encoded as ϛ̄/ϛ̅ alternates — flag.
• Manchu: Unicode correspondence follows the standard model (shared Mongolian letters + Sibe forms + the five Manchu-specific codepoints; foreign kʻ gʻ hʻ as the Todo KA/GAA/HAA codepoints; ts dz ž as Sibe TSA/ZA/ZHA) — documented in the map, flag if a different font convention is preferred. cy/jy digraphs encoded; sy (ᠰᡳ) deliberately not encoded (clashes with native "si").
• Tamazight: year 2012 taken from the PDF metadata (approved by CC:AAM 2012-01-22; the document prints no version). ⴺ carries dot-below + macron-below per the printed table. The document's example "Aɣlyas" includes proper-noun capitalization; the map yields aɣlyas.
• Tamashek: most vernacular cells have no text layer (substitute font); 21 codepoints are text-layer-grounded, the rest follow the Tuareg unification. Long vowels encoded as doubled letters (Tuareg practice; source glyphs are marks with no Unicode encoding) — flag. c assigned ⵞ (YACH) — flag. ă and ḷ marks unencodable (covered by base-letter rules, noted).
• Vai: complete grid encoded (239 syllables + syllabic nasal). The table itself omits the 28 th/dh/sh/zh-series syllables of the Unicode block; recorded in the notes so it is not read as a transcription gap. ꘌ (lengthener) is an instruction, not a romanization value — no rule, noted.

Source-missing (not encodable, no fabrication)

The batch work order also lists N'Ko (nko-Nkoo) and Tod-Oirat (xal) — no source PDFs for either exist in the reference collection, so they are not encoded here.

Verification

• Fresh interscript-ts main checkout: repo parse gate node scripts/parse-isc.mjs (via tsx) → 417/417 .isc files parse (maps/ + libs/, including these 8).
• Embedded tests run through the ts runtime (parseIsc + iscBundledStrategy + transliterate): 221/221 pass.
• Test counts are mechanical table results (letter values and the documents' own examples); multi-char tests are constructed letter sequences, not transcription claims.