Romanian Word Embeddings – SG & FastText (with PCA)

🔍 Overview
This release contains pretrained Word2Vec word embeddings for the Romanian language, trained using:

1. Skip-Gram (SG) and
2. FastText (FT) architectures
   with dimensionality reduction via PCA.

These embeddings are suitable for:

1. Word-level similarity
2. Semantic analogy tasks
3. Input for classic ML models (e.g., classifiers, clustering)
4. Visualization & exploration

PCA was applied to reduce vector size from 300 ➜ 120 for better efficiency and speed.
Happy embedding!

FastText

v1.3

Update README.md

SG

v1.2

Update README.md

CBOW

v1.1

Update README.md

CBOW_300_25_5

v1.0

Update README.md