This document records, by authority, the reference documents in the Interscript project’s reference collection, which is the archive of standards documents against which the maps are built and verified, and the maps in the corpus that implement them. A reference document is listed whether or not the corpus provides a map for it, and each map is named with the reference standard from which it was encoded when that standard is in the collection. The corpus is the authority for what is encoded.
ICS 630-01 is the Intelligence Community standard for foreign-language transliteration. The standard establishes methods for rendering personal names written in non-Roman scripts into the Roman alphabet and delegates the language-specific tables to annexes administered by the Senior Language Authority of each IC element. The project’s reference collection contains 29 documents for the standard: the standard itself, 23 language annexes lettered A to W, three Chinese appendices, and revisions of Annexes J and M.
| Annex | Language | Maps in the corpus |
|---|---|---|
A |
Arabic |
|
B |
Belarusian |
|
C |
Chechen |
|
D |
Chinese |
None |
E |
Georgian |
|
F |
Hindi and Urdu |
|
G |
Kurdish |
None |
H |
Macedonian |
|
I |
Pashto |
|
J |
Persian and Dari |
|
K |
Russian |
|
L |
Serbian |
|
M |
Ukrainian |
|
N |
Tajik |
|
O |
Bulgarian |
|
P |
Azeri |
|
Q |
Kazakh |
|
R |
Kyrgyz |
|
S |
Tatar |
|
T |
Turkmen |
|
U |
Uyghur |
|
V |
Uzbek |
|
W |
Korean |
|
Twenty-one of the 23 language annexes have maps in the corpus. The Chinese annex (D) and the Kurdish annex (G) do not.
Annex D applies to personal names already rendered in Hanyu Pinyin and
standardises that rendering: surname capitalization, omission of the
umlaut on u, spacing, and apostrophes. The annex assumes the pinyin
form and does not supply a lexicon for the conversion of Han characters
to pinyin. Its appendices are Appendix A, the IC transliteration table;
Appendix B, conversion tables from Wade-Giles, Yale, and Mandarin
Phonetic to pinyin; and Appendix C, a list of common Chinese surnames. A
pinyin normalisation map and Latn-to-Latn conversion maps can be
encoded from the annex. A Han-to-pinyin map requires a character
lexicon, and the corpus provides bgnpcgn-zho-Hans-Latn-1979 for that
conversion.
The folder contains documents for Arabic, Belarusian, Bulgarian, Farsi/Persian and Dari, Hindi/Urdu, Korean, Kurdish, Macedonian, Pashto, Serbian, and Ukrainian. The corpus provides ODNI maps for ten of the eleven languages. The Kurdish document is the same table as Annex G and has no map in the corpus. The Chinese annex is in the ICS 630-01 collection.
| Standard | Title | Maps in the corpus |
|---|---|---|
DIN 1460-2:2011 |
Romanization of Cyrillic alphabets, Part 2: non-Slavic languages |
None |
DIN 31635:2011 |
Romanization of the Arabic alphabet for Arabic, Ottoman Turkish, Persian, Kurdish, and Urdu |
None |
DIN 31636:2018 |
Romanization of the Hebrew alphabet |
None |
DIN 32706:2010 |
Romanization of the Armenian alphabet |
None |
DIN 32707:2010 |
Romanization of the Georgian alphabet |
|
DIN 32708:2014 |
Romanization of Japanese |
None |
DIN 33903:2016 |
Romanization of Tamil |
|
DIN 33904:2018 |
Transliteration of Devanagari |
|
Five of the eight DIN standards in the collection have no map in the
corpus. The corpus also contains din-grc-Grek-Latn-31634-2011-t1,
which implements DIN 31634 for Greek; that standard is not in the
collection.
| Standard | Scope | Maps in the corpus |
|---|---|---|
ISO 9:1995 |
Cyrillic to Latin |
|
ISO 9984:1996 |
Georgian to Latin |
|
ISO/TR 11941:1996 |
Korean to Latin, two methods |
|
All three ISO documents in the collection have maps in the corpus. The corpus’s other ISO maps, which implement ISO 233, ISO 843, ISO 3602, ISO 11940, and ISO 15919, have no source document in the collection.
| Standard | Title | Maps in the corpus |
|---|---|---|
BS 2979:1958 |
Transliteration of Cyrillic and Greek characters |
None |
PD 6505:1982 |
Guide to the romanization of Korean |
None |
The BGN/PCGN systems represented in the corpus and the reference tables
that have no map are listed in bgnpcgn-status.adoc in this directory.
The collection contains 16 documents in the 1964 BGN romanization guide and 30 documents in the 1994 USBGN romanization guide. These documents record earlier editions of systems whose current BGN/PCGN sheets are also in the collection, including Hebrew (1962), Burmese (1970), Khmer (1972), and Lao (1966).
The official catalogue of 78 ALA-LC romanization tables is captured in
alalc-roman.html, and differences against the corpus are recorded in
alalc-diff.txt. The collection contains one table document,
alalc/1997/mongolia.pdf. The corpus contains 55 ALA-LC maps, whose
source tables are not in the collection. Encoding the 47 catalog
systems that are not in the corpus requires acquisition of their table
documents. Several of those systems overlap languages represented under
other authorities, including Burmese, Khmer, Lao, Shan, Hebrew, Pashto,
Thai, Urdu, Uyghur, Uzbek, and Kurdish.
The collection contains ICAO 9303 Part 3 (Amendment 1) and the 9303 Arabic helper. The corpus contains eight ICAO maps. The Part 3 in the collection does not contain the language transliteration tables, which are published in another part or edition of 9303.
-
The two BAS documents have maps:
bas-rus-Cyrl-Latn-2017-bssandbas-rus-Cyrl-Latn-2017-oss. -
The corpus contains 32 UN maps, and the UNGEGN schedules are not in the collection.
-
The remaining single-system authorities, including ELOT, MVD, GOST, MEXT, MOCT, LSHK, ACADSIN, SAC, HK, and KP, have verified maps in the corpus and no source document in the collection.
Twenty-five reference tables in the collection have no map in the
corpus: two ICS annexes, five DIN standards, two BSI standards, and
sixteen BGN/PCGN systems. The BGN/PCGN systems are listed in
bgnpcgn-status.adoc; the others are listed in their authority
sections above.
The deterministic map and the neural model are independent products. The corpus can provide a map for a language without a published model, and most languages in the corpus have no model. Neural models are published for Hebrew, Khmer, Thai, Persian, Arabic, and Urdu. No neural model is published for Burmese, Dzongkha, Inuktitut, Lao, or Shan, because no parallel corpus of source-script names with romanized equivalents is available for training in those languages.