visualize.py

import logging
import numpy as np
import matplotlib.pyplot as plt

# from sklearn.manifold import TSNE

from tqdm import tqdm
from gensim.models import Word2Vec
import matplotlib.font_manager as fm
from MulticoreTSNE import MulticoreTSNE as TSNE


# logging
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)

# font
path = '/usr/share/fonts/truetype/nanum/NanumGothic.ttf'  # Korean font is needed!
# path = 'C:\\Windows\\Fonts\\NanumGothic.ttf'
font_name = fm.FontProperties(fname=path, size=16).get_name()
plt.rc('font', family=font_name)


def tsne_plot(model):
    labels, tokens = [], []
    
    for w in model.wv.vocab:
        tokens.append(model[w])
        labels.append(w)
    
    config = {
        'n_jobs': 8,
        'perplexity': 40,
        'n_components': 2,
        'n_iter': 2500,
        'init': 'random',
        'random_state': 42,
        'verbose': 2,
    }
    tsne_model = TSNE(**config)
    
    val = tsne_model.fit_transform(np.asarray(tokens))
    print("[*] t-SNE training done!")

    x = [v[0] for v in tqdm(val)]
    y = [v[1] for v in tqdm(val)]

    plt.figure(figsize=(32, 32))
    plt.title('w2v embeddings vis')

    for i in tqdm(range(len(x))):
        plt.scatter(x[i], y[i]) 
        plt.annotate(labels[i], xy=(x[i], y[i]), xytext=(5, 2),
                     textcoords='offset points', ha='right', va='bottom')

    plt.savefig('tsne-w2v.png')
    plt.show()


if __name__ == '__main__':
    w2v_model = Word2Vec.load('ko_w2v.model')
    print("[+] w2v model loaded!")

    tsne_plot(w2v_model)