aalok-sathe / sanskrit_ipa

Implementation of LREC-CCURL'18 paper: a rule-based system for the transcription of Sanskrit text written using Devanagari into the International Phonetic Alphabet. Uses closest known approximate pronunciations of sounds as well as prosodic and metric rules for syllabification using the WWG algorithm for Sinhalese adapted to Sanskrit and the assignment of syllable-weight-determined stress

sanskrit_IPA

Usage: To start up the application command prompt, do: python3 init.py

Commands:

1. transcribe [text] 1.1: transcribes provided Sanskrit [text] which is written using the Devanagari script into IPA and provides inline response. 1.2: if input file has been set, transcribes text from the input file.

2. setInPath 2.1: sets the input path if user wishes to transcribe text from a particular input file.

sanskrit_ipa's Issues

Executable init.py

Could you make init.py executable (that is, run chmod +x init.py and push the file)? It already has the appropriate shebang at the top and typing ./init.py is preferred to python3 init.py. Also, the name init.py is a bit misleading, because it does not simply initialize something but actually runs it too. How about main.py?

Syllabification issues for words with श्च

These words don't get syllabified:

> transcribe निश्चित
niɕt͡ɕit̪ə
> transcribe निश्चय
niɕt͡ɕəjə
> transcribe नरश्च गजश्च
nəɹəɕt͡ɕə gəd͡ʑəɕt͡ɕə

But this one does:

> transcribe पुनःश्च
pu.ˈnəhɕ.t͡ɕə

In my opinion, the code should reach this condition but doesn't—might be a starting point for debugging.

(Note: at first I thought this was an issue relating to zero-width non-joiners, but it wasn't. Nevertheless, the presense of ZWNJs in the input gives a warning in the script. I believe Sanskrit text does not use ZWNJs but it might be good to explicitly ignore them anyway.)