Pre-trained word vectors for Vietnamese

This project contains embedding for 437,485 Vietnamese unique words from a corpus of 93,546 top documents from Vietnamese Wikipedia. Note that all words are tokenized words.

Requirements

• gensim > =0.13.1 (for Word2Vec)
• pyvi (for word tokenizer)

Background / References

• Check this to know what word embedding is.
• Check this to quickly get a picture of Word2vec.

Pre-trained models

Download the ML4U vi embedding.

Example

In[0]: 
from gensim import models
words = models.KeyedVectors.load_word2vec_format('ML4U_QA_embed.txt', binary=False)
words.most_similar('vua')

Out[5]: 
[('hoàng_đế', 0.8315562009811401),
('quốc_vương', 0.7650110721588135),
('pharaon', 0.699152410030365),
('hoàng_hậu', 0.6941052675247192),
('khalip', 0.6764698028564453),
('hoàng_tử', 0.6755297183990479),
('hoàng_thân', 0.6730912923812866),
('chư_hầu', 0.6716309785842896),
('thái_tử', 0.6599626541137695),
('danh_tướng', 0.6498899459838867)]