Re:Mother Tongues
by chetanahuja on Wednesday May 04, 2016 @03:52AM
(#11626860)
Devnagri script, which is used for a bunch of Indian languages including Hindi, is probably the simplest to transliterate. Each letter has a pretty well defined pronunciation with very rare exceptions. Vowels are completely regular and always have the same sound in all words with no exceptions (that I can think of). It should be pretty easy to write some simple logic in a few hundred lines of code to do this automatically... with the main difficulty being the target language rather than the source language.
[ Parent
| on HN ]
One of my friends once told me, "We don't have spelling bees in India... it wouldn't make sense, when you hear the word there's really only one way to spell it."
I on the other hand can't hear the difference between Hindi "b", "v", "g", and "k" half the time :)
Re:Mother Tongues
by chetanahuja on Thursday May 05, 2016 @01:49AM
(#11634227)
Well the Hindi alphabet has 52 letters with 28 "pure" consonants (which multiply with some modifiers). See here: http://www.omniglot.com/writing/hindi.htm Some other Indian languages have additional consonants that are not familiar to predominantly Hindi speakers. The key thing is that the written alphabet seems to have been designed to accurately represent all of those sounds. It's hard (actually impossible) to map all these to Roman alphabet.