Issue 1362: Add support for sanskrit transliteration in latin/roman script

[jimregan]

https://code.google.com/p/tesseract-ocr/issues/detail?id=1362
(http://web.archive.org/web/20150413012213/https://code.google.com/p/tesseract-ocr/issues/detail?id=1362)

[Shreeshrii]

@theraysmith

The best/Latin.traineddata does not include all the characters that are required for IAST.

I would appreciate if you can either include all characters from this unicharset to Latin traineddata or add this as a separate option.

https://sanskritdocuments.org/iast/by-category/stotra.php
http://sarit.indology.info/

have a number of documents in IAST format.

[Shreeshrii]

Plusminus training worked ok for this (unlike for Devanagari script)

see san_latn.traineddata for version 1

from https://github.com/Shreeshrii/tessdata4alpha/tree/master/best

kamakoti-Latin.txt

kamakoti-san_latn_1.txt

[Shreeshrii]

When testing a page which has both English and IAST(modified transliteration scheme) in it, Latin traineddata gives better results. The traineddata after plusminus tuning does not.

See attached sample.

apracticalhinds01mathgoog_0050-eng+san_latn_1.txt
apracticalhinds01mathgoog_0050-Latin.txt
apracticalhinds01mathgoog_0050-san_latn_1.txt
apracticalhinds01mathgoog_0050-san_latn_2.txt

apracticalhinds01mathgoog_0050-eng.txt