-
Notifications
You must be signed in to change notification settings - Fork 0
/
data_module.py
235 lines (188 loc) · 7.06 KB
/
data_module.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
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
import pickle
import numpy as np
import pandas as pd
import pytorch_lightning as pl
import torch
from config import *
from preprocessing import preprocess
from textblob import TextBlob
# from sklearn.model_selection import train_test_split
from torch.utils.data import DataLoader, Dataset, random_split
from tqdm import tqdm
from transformers import RobertaTokenizer
# Set a random seed for reproducibility
pl.seed_everything(random_state, workers=True)
class SARCDataset(Dataset):
def __init__(self, X, y, tokenizer):
texts = None
if use_sentiment_context:
texts = X[0]
sentiment_polarities = X[1]
sentiment_subjectivities = X[2]
texts = [
preprocess(text, p, s)
for text, p, s in tqdm(
zip(texts, sentiment_polarities, sentiment_subjectivities),
desc="Preprocessing",
total=len(texts),
)
]
else:
texts = X
texts = [preprocess(text) for text in tqdm(texts, desc="Preprocessing")]
# texts = [preprocess(text) for text in tqdm(texts, desc="Preprocessing")]
self._print_random_samples(texts)
self.texts = [
tokenizer(
text,
padding="max_length",
max_length=150,
truncation=True,
return_tensors="pt",
)
for text in tqdm(texts, desc="Tokenizing")
]
self.labels = y
def _print_random_samples(self, texts):
print("Random samples after preprocessing:")
random_entries = np.random.randint(0, len(texts), 5)
for i in random_entries:
print(f"Entry {i}: {texts[i]}")
print()
def __len__(self):
return len(self.texts)
def __getitem__(self, idx):
text = self.texts[idx]
label = -1
if hasattr(self, "labels"):
label = self.labels[idx]
return text, label
class SarcasmDetectionDataModule(pl.LightningDataModule):
def __init__(self, data_file, batch_size=8, num_workers=0, mode="train"):
super().__init__()
self.data_file = data_file
self.batch_size = batch_size
self.num_workers = num_workers
self.sarcasm_df = None
self.mode = mode
self.dataset = None
self.prepare_data()
def prepare_data(self):
print("Preparing data...")
sarcasm_df = pd.read_csv(self.data_file)
# sarcasm_df = pd.read_csv("data/sarcasm_preprocessed_sentiments.csv")
# self.sarcasm_df = sarcasm_df
# return
# We just need comment & label columns
# So, let's remove others.
sarcasm_df.drop(
[
"author",
"subreddit",
"score",
"ups",
"downs",
"date",
"created_utc",
"parent_comment",
],
axis=1,
inplace=True,
)
print("Removing empty rows...")
# remove empty rows
sarcasm_df.dropna(inplace=True)
# Some comments are missing, so we drop the corresponding rows.
sarcasm_df.dropna(subset=["comment"], inplace=True)
# Calculate the lengths of comments
comment_lengths = [len(comment.split()) for comment in sarcasm_df["comment"]]
# Calculate the mean, maximum, and minimum lengths
mean_length = sum(comment_lengths) / len(comment_lengths)
max_length = max(comment_lengths)
min_length = min(comment_lengths)
# Print the results
print("Mean length:", mean_length)
print("Maximum length:", max_length)
print("Minimum length:", min_length)
print("Removing comments with length > 50...")
# Filter the dataframe to keep only comments with length <= 50
mask = [length <= 50 for length in comment_lengths]
sarcasm_df = sarcasm_df[mask]
# Reset the index of the dataframe
sarcasm_df.reset_index(drop=True, inplace=True)
if use_sentiment_context:
print("Adding sentiment columns...")
# Add sentiment polarity and subjectivity columns
sarcasm_df["sentiment_polarity"] = sarcasm_df["comment"].apply(
lambda x: round(TextBlob(x).sentiment.polarity, 1)
)
sarcasm_df["sentiment_subjectivity"] = sarcasm_df["comment"].apply(
lambda x: round(TextBlob(x).sentiment.subjectivity, 1)
)
# Save the dataframe
sarcasm_df.to_csv("data/sarcasm_preprocessed_sentiments.csv", index=False)
self.sarcasm_df = sarcasm_df
def setup(self, stage=None):
print("Setting up data...")
# print("Value counts:", self.sarcasm_df["label"].value_counts())
# X_train, X_test, y_train, y_test = train_test_split(
# self.sarcasm_df["comment"],
# self.sarcasm_df["label"],
# test_size=test_size,
# random_state=random_state,
# )
# train_dataset = SARCDataset(X_train, y_train, tokenizer)
# test_dataset = SARCDataset(X_test, y_test, tokenizer)
tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
if not use_sentiment_context:
self.dataset = SARCDataset(
self.sarcasm_df["comment"], self.sarcasm_df["label"], tokenizer
)
else:
self.dataset = SARCDataset(
(
self.sarcasm_df["comment"],
self.sarcasm_df["sentiment_polarity"],
self.sarcasm_df["sentiment_subjectivity"],
),
self.sarcasm_df["label"],
tokenizer,
)
# save the dataset using torch
# torch.save(self.dataset, "preprocessed_dataset_sentiments.pt")
# load the dataset
# self.dataset = torch.load("preprocessed_dataset_sentiments.pt")
# Split the dataset into train and test set
total_size = len(self.dataset)
train_size = int(TRAIN_SIZE * total_size)
test_size = total_size - train_size
train_dataset, test_dataset = random_split(
self.dataset, [train_size, test_size]
)
self.train_dataset = train_dataset
self.test_dataset = test_dataset
def train_dataloader(self):
return DataLoader(
self.train_dataset,
batch_size=self.batch_size,
shuffle=True,
num_workers=self.num_workers,
)
def val_dataloader(self):
return DataLoader(
self.test_dataset,
batch_size=self.batch_size,
num_workers=self.num_workers,
)
def test_dataloader(self):
return DataLoader(
self.test_dataset,
batch_size=self.batch_size,
num_workers=self.num_workers,
)
def predict_dataloader(self):
return DataLoader(
self.test_dataset,
batch_size=self.batch_size,
num_workers=self.num_workers,
)