-
Notifications
You must be signed in to change notification settings - Fork 1
/
preprocessing.py
207 lines (152 loc) · 6.82 KB
/
preprocessing.py
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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
import data_cleaning
import twint_scraping
import os
from collections import Counter
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split
from torch.utils.data import Dataset, DataLoader
import torch
class config:
'''
Configuration class to store and tune global variables
'''
PAD = '___PAD___'
UNKNOWN = '___UNKNOWN___'
paths = ['./training_data/depressive1.json',
'./training_data/depressive2.json',
'./training_data/depressive3.json',
'./training_data/depressive4.json',
'./training_data/depressive5.json',
'./training_data/depressive6.json',
'./training_data/non-depressive1.json',
'./training_data/non-depressive2.json',
'./training_data/non-depressive3.json',
'./training_data/non-depressive4.json',
'./training_data/non-depressive5.json',
'./training_data/non-depressive6.json']
labels = ['depressive', 'depressive', 'depressive', 'depressive', 'depressive', 'depressive',
'not-depressive', 'not-depressive', 'not-depressive', 'not-depressive',
'not-depressive', 'not-depressive']
save_path = './training_data/all_training_data.csv'
keywords = ['depressed', 'lonely', 'sad', 'depression', 'tired', 'anxious',
'happy', 'joy', 'thankful', 'hope', 'hopeful', 'glad']
nr_of_tweets = [5000, 5000, 5000, 5000, 5000, 5000,
5000, 5000, 5000, 5000, 5000, 5000]
hashtags_to_remove = []
encoder = None
vocab = None
vocab_size = 0
n_classes = 0
def collect_dataset(paths, keywords, nr_of_tweets, hashtags_to_remove, collect=True):
'''
Collecting the dataset and cleans the data
Input: paths - path to where to save the collected tweets (type: list of strings)
keywords - keywords to be used for collecting tweets (type: list of strings)
nr_of_tweets - number of tweets to be collected for each collecting process (type: list of ints)
collect - specifying if to collect tweets or not (type: boolean)
Output: dataset - cleaned dataset of the tweet texts and their labels (type: list if lists)
'''
roots, exts = [], []
for path in paths:
root, ext = os.path.splitext(path)
roots.append(root)
exts.append(ext)
#roots, exts = [os.path.splitext(path) for path in paths]
save_root, save_exts = os.path.splitext(config.save_path)
json_paths = [root+'.json' for root in roots]
csv_path = save_root+'.csv'
if collect:
for idx, json_path in enumerate(json_paths):
twint_scraping.collect_tweets(keywords=keywords[idx], nr_tweets=nr_of_tweets[idx], output_file=json_path)
dataset, keys = data_cleaning.datacleaning(paths=json_paths, labels=config.labels, hashtags_to_remove=hashtags_to_remove,
save_path=csv_path)
return dataset, keys
class DocumentDataset(Dataset):
'''
Basic class for creating dataset from the input and label data
'''
def __init__(self, X, Y):
self.X = X
self.Y = Y
def __getitem__(self, idx):
return self.X[idx], self.Y[idx]
def __len__(self):
return len(self.X)
class DocumentBatcher:
'''
Process the batches to desired output by transform into torch tensors and pads uneven input text data
to the same length
'''
def __init__(self, voc):
self.pad = voc.get_pad_idx()
def __call__(self, XY):
max_len = max(len(x) for x, _ in XY)
Xpadded = torch.as_tensor([x + [self.pad] * (max_len - len(x)) for x, _ in XY])
Y = torch.as_tensor([y for _, y in XY])
return Xpadded, Y
class Vocab:
'''
Encoding the documents
'''
def __init__(self):
# Splitting the tweets into words as tokenizer
self.tokenizer = lambda s: s.split()
def build_vocab(self, docs):
'''
Building the vocabulary from the documents, i.e creating the
word-to-encoding and encoding-to-word dicts
Input: docs - list of all the lines in the corpus
'''
freqs = Counter(w for doc in docs for w in self.tokenizer(doc))
freqs = sorted(((f, w) for w, f in freqs.items()), reverse=True)
self.enc_to_word = [config.PAD, config.UNKNOWN] + [w for _, w in freqs]
self.word_to_enc = {w: i for i, w in enumerate(self.enc_to_word)}
def encode(self, docs):
'''
Encoding the documents
Input: docs - list of all the lines in the corpus
'''
unkn_index = self.word_to_enc[config.UNKNOWN]
return [[self.word_to_enc.get(w, unkn_index) for w in self.tokenizer(doc)] for doc in docs]
def get_unknown_idx(self):
return self.word_to_enc[config.UNKNOWN]
def get_pad_idx(self):
return self.word_to_enc[config.PAD]
def __len__(self):
return len(self.enc_to_word)
def preprocess(batch_size=64, collect=True):
'''
Function for preprocessing the data which splits the data into train/val, builds the vocabulary, fits
the label encoder and creates the dataloaders for the train and validation set
Input: batch_size - batch size to be used in the data loaders (type: int)
collect - specifying if to collect data or not (type: boolean)
Output: dataloaders - the created data loaders for training and validation set (type: list of data loaders)
vocab_size - size of the built vocabulary (type: int)
n_classes - number of classes/ladels in the dataset
'''
data, keys = collect_dataset(paths=config.paths, keywords=config.keywords,
nr_of_tweets=config.nr_of_tweets,
hashtags_to_remove=config.hashtags_to_remove,
collect=collect)
X, Y = data
x_train, x_val, y_train, y_val = train_test_split(X, Y, test_size=0.2, shuffle=True, random_state=1)
vocab = Vocab()
vocab.build_vocab(x_train)
config.vocab = vocab
encoder = LabelEncoder()
encoder.fit(y_train)
config.encoder = encoder
vocab_size = len(vocab)
n_classes = len(encoder.classes_)
config.vocab_size = vocab_size
config.n_classes = n_classes
batcher = DocumentBatcher(vocab)
train_dataset = DocumentDataset(vocab.encode(x_train),
encoder.transform(y_train))
train_loader = DataLoader(train_dataset, batch_size,
shuffle=True, collate_fn=batcher, drop_last=True)
val_dataset = DocumentDataset(vocab.encode(x_val), encoder.transform(y_val))
val_loader = DataLoader(val_dataset, batch_size,
shuffle=True, collate_fn=batcher, drop_last=True)
dataloaders = [train_loader, val_loader]
return dataloaders, vocab_size, n_classes