/
classifier.py
185 lines (148 loc) · 6.23 KB
/
classifier.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
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score
from sklearn.metrics import classification_report
from sklearn.feature_extraction import DictVectorizer
from operator import itemgetter
from nltk.stem import PorterStemmer
from nltk.tokenize import word_tokenize
import nltk
import pandas as pd
import numpy as np
import os, re
from nltk.corpus import stopwords
import annotation
STOPWORDS = set(stopwords.words('english'))
def clean_text(text):
text = re.sub('<[^>]+>', '', text) # erase HTML tags
text = re.sub(r'[^A-Za-z]+', ' ', text)
return text.lower()
def stem_tokens(stemmer, words):
# text = re.sub(r'\d+', '', text)
words_without_stopwords = []
for word in words:
if word not in STOPWORDS:
# print(word)
words_without_stopwords.append(word)
return [stemmer.stem(token) for token in words_without_stopwords]
# In[206]:
def bigrams(words):
bigrams = list(nltk.bigrams(words))
bigrams_lower = [(bigram1.lower(), bigram2.lower()) for (bigram1, bigram2) in bigrams]
bigram_not_stopwords = []
for bigram in bigrams_lower:
# remove stopwords bigrams
if bigram[0] not in STOPWORDS or bigram[1] not in STOPWORDS:
bigram_not_stopwords.append(bigram)
return bigram_not_stopwords
# return word_tokenize(text)
# In[207]:
def trigrams(words):
trigrams = list(nltk.trigrams(words))
trigrams_lower = [(trigram1.lower(), trigram2.lower(), trigram3.lower()) for (trigram1, trigram2, trigram3) in
trigrams]
trigram_not_stopwords = []
for trigram in trigrams_lower:
# remove stopwords bigrams
if trigram[0] not in STOPWORDS and trigram[2] not in STOPWORDS:
trigram_not_stopwords.append(trigram)
return trigram_not_stopwords
# return word_tokenize(text)
def extract_words(path_to_file, stemmer):
with open(path_to_file, 'r') as file:
list_words = file.read().splitlines()
list_words_stemmed = [stemmer.stem(token) for token in list_words]
# print(list_words)
return list_words_stemmed
def extract_feat_vocab(file):
data_frame = pd.read_csv(file, encoding='latin1')
feat_vocab = dict()
for index, row in data_frame[data_frame['type'] == 'train'].iterrows():
text = clean_text(row['tweet'])
tokens = word_tokenize(text)
for token in tokens:
feat_vocab[token] = feat_vocab.get(token, 0) + 1
return feat_vocab
def select_features(feat_vocab, most_freq=100, least_freq=5000):
sorted_feat_vocab = sorted(feat_vocab.items(), key=itemgetter(1), reverse=True)
feat_dict = dict(sorted_feat_vocab[most_freq:len(sorted_feat_vocab) - least_freq])
return set(feat_dict.keys())
"""
feat_dict = dict()
for key, value in feat_vocab.items():
if value>1:
feat_dict[key] = value
return set(feat_dict.keys())
"""
def featurize(csv_file, feat_vocab):
cols = ['_type_', '_label_']
cols.extend(list(feat_vocab))
#cols.extend(['positive_words', 'negative_words', 'subjectivity_words'])
data_frame = pd.read_csv(csv_file, encoding='latin1', engine='python')
# dtype={'FULL': 'str', 'COUNT': 'int'}
row_count = data_frame.shape[0]
feat_data_frame = pd.DataFrame(index=np.arange(row_count), columns=cols)
feat_data_frame.fillna(0, inplace=True) # inplace: mutable
for index, row in data_frame.iterrows():
feat_data_frame.loc[index, '_type_'] = row['type']
feat_data_frame.loc[index, '_label_'] = row['label']
text = clean_text(row['tweet'])
tokens = word_tokenize(text)
for token in tokens:
if token.lower() in feat_vocab:
feat_data_frame.loc[index, token] += 1
return feat_data_frame
def vectorize(feature_csv, split="train"):
df = pd.read_csv(feature_csv, encoding='latin1', engine='python')
df = df[df['_type_'] == split]
df.fillna(0, inplace=True)
data = list()
for index, row in df.iterrows():
datum = dict()
datum['bias'] = 1
for col in df.columns:
if not (col == "_type_" or col == "_label_" or col == 'index'):
datum[col] = row[col]
data.append(datum)
#print(data)
vec = DictVectorizer()
data = vec.fit_transform(data).toarray()
#print(data.shape)
labels = df._label_.as_matrix()
#print(labels.shape)
return data, labels
def train_model(X_train, y_train, model):
model.fit(X_train, y_train)
print("Shape of model coefficients and intercepts: {} {}".format(model.coef_.shape, model.intercept_.shape))
return model
def test_model(X_test, y_test, model):
predictions = model.predict(X_test)
report = classification_report(predictions, y_test)
accuracy = accuracy_score(predictions, y_test)
return accuracy, report
def classify(feat_csv):
X_train, y_train = vectorize(feat_csv)
X_test, y_test = vectorize(feat_csv, split='test')
model = LogisticRegression(multi_class='multinomial', penalty='l2', solver='lbfgs', max_iter=20,
verbose=1)
model = train_model(X_train, y_train, model)
accuracy, report = test_model(X_test, y_test, model)
print(report)
def select_rows(from_csv, to_csvfile):
from_df = pd.read_csv(from_csv, encoding='latin1', engine='python')
to_df = pd.DataFrame(columns=['type', 'tweet', 'label'])
to_index = 0
for index, row in from_df.iterrows():
if index % 1 == 0: # 10 original
to_df.loc[to_index] = [row['type'], row['review'], row['label']]
to_index += 1
to_df.to_csv(to_csvfile, encoding='latin1')
# In[217]:
if __name__ == '__main__':
#replace filename to run on different data set
file = "sarcasm vs everything.csv"
feat_vocab = extract_feat_vocab(file)
selected_feat_vocab = select_features(feat_vocab, 1, 50)
feat_data_frame = featurize(file, selected_feat_vocab)
featfile = os.path.join(os.path.curdir, "features.csv")
feat_data_frame.to_csv(featfile, encoding='latin1', index=False)
classify('features.csv')