Vreme: 11/17/2011 03:29 PM, Rajagopal Swaminathan piše:
OTOH, Voice training is easier using the language substrate of Sanskrit (the oldest and most well defined "language" spoken by very few in India) and can be easily adapted to Indian Languages, at least.
Actually, simplest Voice recognition would be for ex-Yugoslavia (Serbian-Croatian, Slovenian and Macedonian) languages. The language reform in late 19th century gave us "one sound = one letter, one letter = one sound" reading/writing. There are 30 sounds and letters in Cyrillic and 27+3 in Latin script.
So "voice" would be written as "vojs", "people" as "pipl", "trouble" as "trabl", "throughout" as "truaut" and "throughput" as "truput".
Grammar is somewhat harder, but as for voice recognition goes, I have never heard of easier language.
But there is only 20-30 million speaking it/them.