ReviewGraph/word2vector.py at main · JarenceSJ/ReviewGraph · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
import os
from collections import defaultdict
from load_data import load_corpus, get_dir_and_base_name
import json
import numpy as np
from gensim.test.utils import datapath, get_tmpfile
from gensim.models import KeyedVectors
from gensim.scripts.glove2word2vec import glove2word2vec

GLOVE_PATH = '/home/d1/shuaijie/data/glove.6B.100d.txt'
GLOVE_TMP_PATH = '/home/d1/shuaijie/data/glove.6B.100d.word2vec.txt'

# GLOVE_PATH = 'f:/data/glove.6B.100d.txt'
# GLOVE_TMP_PATH = 'f:/data/glove.6B.100d.word2vec.txt'


class Args:
    # dataset_name = 'Office_Products_5'
    # dataset_path = '/shuaijie/data/Office_Products_5/Office_Products_5.json'

    # dataset_name = 'Instant_Video_5'
    # dataset_path = '/home/d1/shuaijie/data/Instant_Video_5/Instant_Video_5.json'

    # dataset_name = 'Digital_Music_5'
    # dataset_path = '/home/d1/shuaijie/data/Digital_Music_5/Digital_Music_5.json'

    # dataset_name = 'Sports_and_Outdoors_5'
    # dataset_path = '/home/d1/shuaijie/data/Sports_and_Outdoors_5/Sports_and_Outdoors_5.json'

    dataset_name = 'Clothing_5'
    # dataset_path = '/home/d1/shuaijie/data/Clothing_5/Clothing_5.json'
    dataset_path = '/home/d1/shuaijie/NeuralEDUSeg/data/Clothing_5/Clothing_5.json'

    # dataset_name = 'Toys_and_Games_5'
    # dataset_path = '/home/d1/shuaijie/data/Toys_and_Games_5/Toys_and_Games_5.json'

    # dataset_name = 'Health_and_Personal_Care_5'
    # dataset_path = '/home/d1/shuaijie/data/Health_and_Personal_Care_5/Health_and_Personal_Care_5.json'

    # dataset_name = 'CDs_and_Vinyl_5'
    # dataset_path = '/home/d1/shuaijie/data/CDs_and_Vinyl_5/CDs_and_Vinyl_5.json'

    # dataset_name = 'Movies_and_TV_5'
    # dataset_path = '/home/d1/shuaijie/data/Movies_and_TV_5/Movies_and_TV_5.json'

    # dataset_name = 'Baby_5'
    # dataset_path = '/home/d1/shuaijie/data/Baby_5/Baby_5.json'

    # dataset_name = 'Yelp2013'
    # dataset_path = '/home/d1/shuaijie/data/yelp-recsys-2013/yelp2013.json'

    vocab_size = 50000
    embedding_dim = 100

    args_str = 'embed_dim_{}'.format(embedding_dim)


dir_path, _ = get_dir_and_base_name(Args.dataset_path)

Args.word2id_path = \
    '{}/word2id_{}.json'.format(dir_path, Args.args_str)
Args.embedding_path = \
    '{}/word_embedding_{}.npy'.format(dir_path, Args.args_str)


def load_glove():
    tmp_file = get_tmpfile(GLOVE_TMP_PATH)

    if not os.path.exists(GLOVE_TMP_PATH):
        glove_file = datapath(GLOVE_PATH)

        _ = glove2word2vec(glove_file, tmp_file)

    model = KeyedVectors.load_word2vec_format(tmp_file)

    # word2id = {word: index for index, word in enumerate(model.index2word)}
    word2id = model.key_to_index
    embed = model.vectors

    return word2id, embed


if __name__ == '__main__':
    args = Args
    print('Load Word2Vec from Glove, on dataset {}'.format(args.dataset_name))

    sentences = load_corpus(args.dataset_path)
    sentences = [x.split() for x in sentences]

    corpus_word_num = 0
    corpus_word_counter = defaultdict(int)
    for words in sentences:
        corpus_word_num += len(words)
        for word in words:
            corpus_word_counter[word] += 1

    corpus_word_counter = sorted(corpus_word_counter.items(),
                                 key=lambda x: x[1], reverse=True)
    corpus_word_counter = corpus_word_counter[:args.vocab_size]
    corpus_word_counter = dict(corpus_word_counter)

    glove_word2id, glove_embed = load_glove()
    corpus_word_set = set(corpus_word_counter.keys())
    glove_word_set = set(glove_word2id.keys())

    used_words = list(corpus_word_set & glove_word_set)
    unused_words = list(corpus_word_set - glove_word_set)

    unused_words_counter = {k: corpus_word_counter[k] for k in unused_words}
    unused_words_counter = dict(sorted(unused_words_counter.items(),
                                       key=lambda x: x[1], reverse=True))

    ds_word2id = {'<PAD>': 0}
    ds_embed = list()
    ds_embed.append(np.zeros(glove_embed.shape[1]))

    # used words
    for w in used_words:
        ds_word2id[w] = len(ds_word2id)
        ds_embed.append(glove_embed[glove_word2id[w]])

    for w in unused_words:
        ds_word2id[w] = len(ds_word2id)
        ds_embed.append(np.random.normal(.0, 1., glove_embed.shape[1]))

    ds_embed = np.stack(ds_embed, axis=0)

    with open(args.word2id_path, 'w') as f:
        json.dump(ds_word2id, f)

    np.save(args.embedding_path, ds_embed)