natasha/navec

Compact high quality word embeddings for Russian language

51
/ 100
Established

Trained on large Russian corpora (12B+ tokens) using vanilla GloVe with vector quantization (300d→100d), achieving competitive intrinsic evaluation scores while reducing disk footprint to ~50MB and load time to ~1 second. Integrates directly with PyTorch via the Slovnet library's `NavecEmbedding` module and supports dictionary-like access patterns for both word lookup and vocabulary indexing with special handling for unknown and padding tokens.

216 stars. Used by 2 other packages. No commits in the last 6 months. Available on PyPI.

Stale 6m
Maintenance 0 / 25
Adoption 12 / 25
Maturity 25 / 25
Community 14 / 25

How are scores calculated?

Stars

216

Forks

20

Language

Python

License

MIT

Last pushed

Jul 24, 2023

Commits (30d)

0

Dependencies

1

Reverse dependents

2

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/embeddings/natasha/navec"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.