lat-koi — Reverse transliteration

• Get the sources here: lat-koi and lat-koi-ukr.

• lat-koi is a Perl script that converts transliterated Russian text into KOI-8 (i.e. does reverse transliteration). lat-koi-ukr does the same for Ukrainian text.

• You can use these scripts on your own system. Source text has to be given on standard input and converted text appears on standard output. So if you want, say, to reverse transliterate file-lat into file-koi in KOI8, the command is

  lat-koi < file-lat > file-koi 

• It should be understood that reverse transliteration is an intrinsically ambiguous operation, because certain Cyrillic characters correspond to two, or even more, Latin ones. (There do exist schemes with one-to-one correspondence, where the "problematic" Cyrillic characters are displayed by brackets etc. — but as for me, I can't read them and have no desire to support them.) There is no way, short of teaching complete (or almost complete) words to the program, to let it know that the ts in detskij and molodetskij means two different things. And with pasha, you are stuck altogether.

• This might be resolved twofold: either by avoiding the ambiguous combinations where possible: molodeckij and paskha vs pasha; or by inserting a tilde (~) in the middle of those which need to be rendered as two Cyrillic characters: det~skij, pas~ha.

• That said, the program does know many common words; therefore, no special tricks will usually be required. It being all but impossible to devise a program that will do it 100% right, this one will have a perhaps 99.9% success rate (with Russian) unless you give it some really weird transliteration. Still, sometimes it will certainly fail. Let's say, it will not always undo the transliteration by cyr-conv correct. Please report failures to me, so I could teach it better!

• The algorithm it uses is exceedingly simple. As it moves on through source text, it checks whether the next N characters make a piece that it knows — and transliterates that piece if it does. If it does not, it decreases N by one and tries again. If even N=1 fails, it must be a space or punctuation sign, which it leaves as is.

• Some examples of how the transliterated words may look for the program to get them correct (where two variants are given, intermediate ones will also work): jakhont or yahont; schiptsy or shchipcy; delaetsya or delajetsja; shveya; zhest'; schast'ye; zajavlyayu or zayavlyaju; pod'yom or pod'em if you don't care about the yo.

• Here is the full transliteration table in KOI-8. [Note: Small soft sign is ' (single quote), capital soft sign is ^' (caret - single quote). Small hard sign is '' (two single quotes — NOT double quote), capital hard sign is ^'' (caret - two single quotes).]

  Char.	Rus.	Ukr.		Char.	Rus.	Ukr.		Char.	Rus.	Ukr.
  А	A	A		Ї		YI,JI		Х	H,KH	H,KH
  Б	B	B		Й	J	J		Ц	C,TS	C,TS
  В	V	V		К	K	K		Ч	CH	CH
  Г	G	G		Л	L	L		Ш	SH	SH
  Г		G		М	M	M		Щ	SCH,SHCH	SCH,SHCH
  Д	D	D		Н	N	N		Ъ	'',^''
  Е	E	E		О	O	O		Ы	Y
  Є		YE,JE		П	P	P		Ь	',^'	',^'
  Ё	YO,JO			Р	R	R		Э	`E
  Ж	ZH	ZH		С	S	S		Ю	YU,JU	YU,JU
  З	Z	Z		Т	T	T		Я	YA,JA	YA,JA
  И	I	Y		У	U	U		'		''
  I		I		Ф	F	F

  Note that this table does not always coincide with those adopted by some standard committees. This is intentional — it is constructed so as to minimize the ambiguities.

• You might also be interested in a good C program program transkoi.c by Ilya Sandler that does the same thing. (And you are welcome to compare the length of code here and there ;-)

• Comments, suggestions, problems? Please email me!