DiacNet1
Collection
A collection of models for diacritic restoration. Explore cutting-edge tools for processing and enhancing text with diacritics. • 8 items • Updated
DiacNetYor-Viterbi
DiacNetYor-Viterbi is a lightweight, statistical full tonal and dot-below diacritization model for Yoruba (yo) text. It uses a bigram Viterbi decoder over a word-level candidate map.
Model Details
- Model Type: Word-level Bigram Viterbi Decoder
- File Size: 7.17 MB (
yoruba_diacritizer.json) - Supported Languages: Yoruba (
yo) - Metrics:
- Word Accuracy: 71.66% (evaluated on 54,815 words)
- Dependencies: None (pure Python / zero dependencies)
Usage
Loaded and used via the unified olaverse SDK wrapper:
from olaverse.nlp.diacritizer import Diacritizer
diacritizer = Diacritizer(model="diacnet-yor-viterbi")
text = "Ojo lo si oja lana"
print(diacritizer.restore(text))
# Output: "Ọjọ́ ló sí ọjà lànà"
Links
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support