basic_LM/data_loader.py at master · Dutil/basic_LM · 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
132
133
134
135
136
137
138
139
140
141
142
143
144
145
from collections import Counter
import numpy as np
import ipdb, utils


class predict_next_iterator:

    def __init__(self, iterator):
        self.iterator = iterator
        self.v_size = len(iterator.vocab)


    def __iter__(self):

        for minibatch in self.iterator:
            minibatch = utils.hotify_minibatch(minibatch, self.v_size)
            yield minibatch[:-1], minibatch[1:]

class predict_noisy_self:
    def __init__(self, iterator, p=0.0):
        self.iterator = iterator
        self.v_size = len(iterator.vocab)
        self.p = p


    def __iter__(self):

        for minibatch in self.iterator:

            noisy_minibatch = [filter(lambda x: np.random.rand() >= self.p, m) for m in minibatch]

            hot_minibatch = utils.hotify_minibatch(minibatch, self.v_size, pad_before=0)
            hot_noisy_minibatch = utils.hotify_minibatch(noisy_minibatch, self.v_size, pad_before=0)


            yield hot_noisy_minibatch, hot_minibatch

class data_iterator:

    def __init__(self, data, e_size, m_size, vocab, wordMapping):

        self.data = data.split("\n")
        np.random.seed(1993)
        np.random.shuffle(self.data)

        self.m_size = m_size
        self.vocab = vocab
        self.wordMapping = wordMapping
        self.e_size = e_size

        self.noEx = 0 # The no of the exemple we are presenty at
        self.nbMinibatch = 0 # The number of minibatch we have done right now

    def __iter__(self):

        minibatch = []
        for i in self.data[self.noEx:]:

            if i: # for empty sentences
                minibatch.append(self.switchRep(i.split()))

            if len(minibatch) >= self.m_size:

                self.noEx += len(minibatch)
                self.nbMinibatch += 1

                yield minibatch
                minibatch = []

            #If we consider that we have done one epoch
            if self.nbMinibatch == self.e_size:
                self.nbMinibatch = 0
                raise StopIteration

        #Left over exemples
        if minibatch:
            yield minibatch

        self.noEx = 0
        self.nbMinibatch = 0

    def switchRep(self, ids):
        sentence = [self.wordMapping[x] if not type(x) == np.ndarray
                else self.wordMapping[np.where(x)[0][0]] if sum(x) > 0 else '' for x in ids]

        return filter(None, sentence)

    def __getitem__(self, key):
        return self.switchRep(self.data[key].split())

class data_crawler:

    def __init__(self, folder="testing_data", maxCount = 10):

        self.folder = folder
        self.maxCount = maxCount

        training_set = open("{}/train.txt".format(folder)).read()
        valid_set = open("{}/valid.txt".format(folder)).read()
        test_set = open("{}/test.txt".format(folder)).read()
        self.all_data = [training_set, valid_set, test_set]

        self._initVocab(" ".join(self.all_data))


    def _initVocab(self, data):

        vocab = Counter()
        wordMapping = {}
        nbWords = 0

        #Get the vocab
        #data = data.replace("\n", "--EOS--")
        vocab.update(data.split())

        #The words we are keeping
        mostCommon = Counter(dict(vocab.most_common(self.maxCount - 1)))
        lessCommon = vocab - mostCommon


        #get the words ids
        for i, word in enumerate(mostCommon):
            wordMapping[word] = i
            wordMapping[i] = word
            nbWords = i + 1

        #The OOV
        #mostCommon["--OOV--"] = sum(lessCommon.values())
        #wordMapping["--OOV--"] = nbWords
        #wordMapping[nbWords] = "--OOV--"

        self.vocab = mostCommon
        self.wordMapping = wordMapping
        self.nbWords = nbWords #+ 1

    def replaceOOV(self, data):

        # replace the less frequent words to OOV
        data = data.split()
        for noWord, word in enumerate(data):
            if word not in self.vocab:
                data[noWord] = "--OOV--"
        data = " ".join(data)
        return data