Word2vec/

polyglot_words_embeddings_hat

import pickle
from numpy import dot
from numpy.linalg import norm
from huggingface_hub import hf_hub_download
words, embeddings = pickle.load(open(hf_hub_download(repo_id="Word2vec/polyglot_words_embeddings_en", filename="words_embeddings_en.pkl"), 'rb'),encoding="latin1")

word = "Irish"
a = embeddings[words.index(word)]
most_similar = []
for i in range(len(embeddings)):
    if i != words.index(word):
        b = embeddings[i]
        cos_sim = dot(a, b)/(norm(a)*norm(b))
        most_similar.append(cos_sim)
    else:
        most_similar.append(0)        
        
words[most_similar.index(max(most_similar))]

@InProceedings{polyglot:2013:ACL-CoNLL,
 author    = {Al-Rfou, Rami  and  Perozzi, Bryan  and  Skiena, Steven},
 title     = {Polyglot: Distributed Word Representations for Multilingual NLP},
 booktitle = {Proceedings of the Seventeenth Conference on Computational Natural Language Learning},
 month     = {August},
 year      = {2013},
 address   = {Sofia, Bulgaria},
 publisher = {Association for Computational Linguistics},
 pages     = {183--192},
 url       = {http://www.aclweb.org/anthology/W13-3520}
}