g2p research

[thewh1teagle]

1. Find exactly what are the used vowels from utf-8
2. Normalize rarely used
3. Simplify rules
4. Test against https://github.com/elazarg/hebrew_diacritized both at letter + vowels and in word level! https://github.com/thewh1teagle/espeak-ng-studio inspire from https://github.com/espeak-ng/espeak-ng/blob/master/dictsource/ar_rules
5. Fix phonemizer to cache espeak-ng instance
6. Check nakdimon and dicta possible letters and vowel points
7. New diacritizer based on https://dumps.wikimedia.org/hewiki with T5 T5G2P

• https://github.com/hexgrad/misaki
• https://en.wikipedia.org/wiki/Unicode_and_HTML_for_the_Hebrew_alphabet#Compact_table
• https://github.com/espeak-ng/espeak-ng/blob/master/dictsource/he_rules
• https://github.com/mush42/hareef/tree/main