Words in similar contexts have similar meanings. For example, the word puppy usually comes with words such as cute, pretty, cuteness.
| CBOW(Continuous Bag of Words) predicts words in the middle by inputting words around the middle. w1 w2 w3...ㅁ...w5 w6 w7 → → → ↑ ← ← ← |
| Skip-Gram predicts surrounding words by inputting words in the middle. ㅁㅁㅁ...w...ㅁㅁㅁ ←←← ↓ →→→ |
- Training Word2Vec model
model=Word2Vec(sentences=result, vector_size=150, window=3, min_count=5, workers=3, sg=0)vector_size : The number of dimensions of each word. How many vector dimensions of one word represents.
min_ count : Limit the minimum frequency of words. Less frequency are not learned.
workers : How many process are you going to use.
sg : Skip-Grams, 0 is CBOW, 1 is Skip-gram.
- Find the most similar word
print(model.wv.most_similar("only"))wv : Word Vectors
- Save the Word2Vec model
model.wv.save_word2vec_format('eng_w2v')
- Load the Word2Vec model
loaded_model=KeyedVectors.load_word2vec_format('eng_w2v')'Deep Learning' 카테고리의 다른 글
| BPE(Byte Pair Encoding) (0) | 2022.11.30 |
|---|---|
| Terms-text encoding, text decoding, embedding (0) | 2022.09.23 |
| How to learning of DL (0) | 2022.03.17 |
| Activation function (0) | 2022.03.17 |
| sklearn (0) | 2022.03.08 |