RenikudPlus — Hebrew grapheme-to-phoneme

Hebrew text in, IPA out, with stress marks — the part unpointed Hebrew leaves to context. For TTS front-ends, lexicon work and pronunciation research.

Results

Target-word accuracy on 1,653 sentences of held-out Hebrew:

	RenikudPlus	Gemini	ReNikud (paper)	Phonikud
OVERALL	88.6%	84.6%	78.2%	66.4%
Gender	99.3%	77.0%	59.9%	37.5%
Min. Stress Pairs	93.3%	88.0%	80.7%	78.7%
Stress Homographs	92.6%	92.6%	79.8%	78.3%
Names	80.0%	75.3%	67.3%	68.0%
Acronyms	82.2%	79.0%	56.6%	37.5%
Slang	82.7%	71.2%	59.0%	41.7%
Penultimate Stress	88.1%	89.4%	82.1%	60.9%
Rare Phonemes	76.8%	57.6%	41.1%	19.9%
Foreign	87.1%	71.6%	56.8%	36.1%
Colloquial	51.0%	25.2%	54.3%	9.3%
ILSpeech-test	93.7%	96.0%	92.5%	85.5%

Usage

pip install onnxruntime numpy

Download renikud_onnx.py and a model file from this repo into the same folder:

from renikud_onnx import G2P

g2p = G2P("model_int8.onnx")

g2p.phonemize("שלום, מה נשמע?")
# ʃalˈom, mˈa niʃmˈa?

g2p.phonemize("הלכתי לספר וקראתי ספר בזמן שהוא ספר כמה אצבעות יש לו")
# halˈaχti lasapˈaʁ vekaʁˈati sˈefeʁ bizmˈan ʃehˈu safˈaʁ kˈama ʔetsbaʔˈot jˈeʃ lˈo
#            barber        book                    counted

Four occurrences of ספר, four different words, each resolved from context.

Speaker and addressee

Hebrew inflects for who is speaking and who is addressed; the text usually shows neither. Both controls default to 0 (unknown), 1 = male, 2 = female.

g2p.phonemize("אני רוצה להגיד לך משהו חשוב", speaker=2, target_speaker=1)

speaking → addressing	אני רוצה להגיד לך משהו חשוב
man → man	ʔanˈi ʁotsˈe lehaɡˈid leχˈa mˈaʃu χaʃˈuv
man → woman	ʔanˈi ʁotsˈe lehaɡˈid lˈaχ mˈaʃu χaʃˈuv
woman → man	ʔanˈi ʁotsˈa lehaɡˈid leχˈa mˈaʃu χaʃˈuv
woman → woman	ʔanˈi ʁotsˈa lehaɡˈid lˈaχ mˈaʃu χaʃˈuv

speaker sets רוצה, target_speaker sets לך. Four readings of one unchanged sentence.

Numbers and digits

Digits are expanded to Hebrew words before phonemization, with gender agreement taken from the counted noun:

g2p.phonemize("יש לי 3 ילדים")        # jˈeʃ lˈi ʃloʃˈa jeladˈim
g2p.phonemize("יש לי 3 בנות")         # jˈeʃ lˈi ʃalˈoʃ banˈot
g2p.phonemize("המחיר 1250 שקלים")     # hameχˈiʁ ʔˈelef matˈajim veχamiʃˈim ʃkalˈim

Files

file	what
model.onnx	fp32
model_int8.onnx	quantized, 4× smaller, faster on CPU
renikud_onnx.py	wrapper — decoding, long-input windowing, number expansion

Notes

• Stress is a first-class output (ˈ before the stressed syllable).
• Remaining errors concentrate in unattested foreign names and in transcription-convention choices (glottal-stop realisation, ba/be proclitics) where more than one reading is correct Hebrew.