I took Sanskrit at the Far Eastern and Slavic Institute while an undergraduate at the University of Washington. Six linguists and me – the hard sciences major. They didn’t know what to make of me.
It immediately struck me that if there ever was language “invented” by/for a computer, it was Sanskrit. I was only 19, so with a few gray hairs, I understand some of the challenges a bit better.
With only seven students in a University of 33,000 taking this class, you can guess that Sanskrit is not a big draw. For those who only know a little about it, let me state a few things. Sanskrit is the ancient language of India. Some claim it is older than Hebrew, but I’ll leave that for the other six students in the class to debate. But it is old.
English comes from the roots where Sanskrit does, but what does it have to do with speech recognition and computers?
Sanskrit is written in an alphabet called the Devanagari and it looks exotic – the writing we see in India, It has an interesting feature. While in today’s India, words are separated by spaces, in ancient Sanskrit all words run together.
If one word ends with a vowel, and the next word begins with a vowel, a “diphthong” is formed. Diphthongs in English are within words. We do think of them as joining words, at least not consciously.
Let me illustrate an example of a diphthong. The English word “meant” has two vowels “ea” which together do not sound like either vowel by itself. It is pronounced in General American as a short “eh” rather than “e-a” as in “me ant,” implying “I live inside a hill with other insects.”
Combining the vowels is called vowel sandhi which makes the language a bit daunting.
Thus, not only do the words run together, when they do, new letters (and sounds) are formed.
Yet, for all that, the language is rather orderly and the letters are laid down matrix-like.
It is a strange notion to me, at least, that a language would be written as it sounds, literally. Maybe the computers would like it better than humans.