Word2Vec_week_2.md

Word2Vec_week_2.md

Word2Vec_ practice

I'm going to practice on Word2Vec using gensim.

gensim

Gensim is an open-source library for unsupervised topic modeling and natural language processing, using modern statistical machine learning.

https://en.wikipedia.org/wiki/Gensim

practice

I referred to this blog. Language is Korean.

https://m.blog.naver.com/PostView.nhn?blogId=wideeyed&logNo=221349385092&proxyReferer=https%3A%2F%2Fwww.google.com%2F

from gensim.models import Word2Vec
import matplotlib.pyplot as plt

#this is a function that draws 2 dimension graph.

def plot_2d_graph(vocabs, xs,ys):
    plt.figure(figsize=(8,6))
    plt.scatter(xs,ys,marker='o')
    for i, v in enumerate(vocabs):
        plt.annotate(v,xy=(xs[i],ys[i]))

sentences = [
                ['this','is','a','good','product'],
                ['it','is','a','excellent','product'],
                ['it','is','a','bad','product'],
                ['that','is','the','worst','product'],
                ['you','are','beautiful'],
                ['you','are','nice'],
                ['he','is','nice'],
                ['she','is','beautiful'],
                ['we','are','good','team'],
                ['you','are','lovely'],
                ['you','are','pretty']
]

#It generates words and vectors from using sentences.

model = Word2Vec(sentences, size=300, window=3, min_count=1, workers=1)

#Get word vectors.

word_vectors=model.wv

vocabs = word_vectors.vocab.keys()
word_vectors_list = [word_vectors[v] for v in vocabs]

#Get similarity between 'this' and 'that'.

print(word_vectors.similarity(w1='this',w2='that'))

#The similarity between 'this' and 'that'.

from sklearn.decomposition import PCA
pca = PCA(n_components=2)
xys=pca.fit_transform(word_vectors_list)
xs = xys[:,0]
ys = xys[:,1]

plot_2d_graph(vocabs,xs,ys)

#The plot shows vectored words.

#Merge it with the model that's already learned.

file_name = 'Downloads/GoogleNews-vectors-negative300.bin'
model.intersect_word2vec_format(fname=file_name, binary=True)

#Get word vectors.

word_vectors = model.wv

vocabs = word_vectors.vocab.keys()
word_vectors_list = [word_vectors[v] for v in vocabs]

#The similarity between 'this' and 'that'.

print(word_vectors.similarity(w1='this',w2='that'))

#We can show improvement of similarity between 'this' ans 'that'.

pca = PCA(n_components=2)
xys=pca.fit_transform(word_vectors_list)
xs = xys[:,0]
ys = xys[:,1]

#We can show improvement of similarity between 'this' ans 'that'.

plot_2d_graph(vocabs,xs,ys)

#We can show improvement of accuracy.