-
Notifications
You must be signed in to change notification settings - Fork 0
/
aspect_extract.py
260 lines (211 loc) · 9.51 KB
/
aspect_extract.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
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
from argparse import ArgumentParser, FileType
from collections import Counter
import argparse
import numpy as np
import pandas as pd
import regex as re
import string
import os
import json
import datetime
from nltk import pos_tag, RegexpParser
from nltk.corpus import stopwords
from nltk.stem.snowball import SnowballStemmer
from nltk.tokenize import sent_tokenize, word_tokenize, TweetTokenizer
from nltk.tokenize.treebank import MacIntyreContractions, TreebankWordTokenizer
from nltk.tokenize.casual import EMOTICON_RE, _replace_html_entities
import matplotlib
matplotlib.use('TkAgg')
import matplotlib.pyplot as plt
import seaborn as sns
# %matplotlib inline
plt.style.use('seaborn-whitegrid')
params = {'figure.figsize': (15,12),
'savefig.facecolor': 'white'}
plt.rcParams.update(params)
pd.set_option('display.max_colwidth', -1)
pd.set_option('display.max_columns', 10)
DEFAULT_DATA_FILE = "./data/sample_data.json"
IMAGES_DIRECTORY = './images'
DEFAULT_SEED = 42
STOPWORDS = set(stopwords.words('english') + ["'ve", "'d", "'s", "one", "use", "would", "get", "also"]) - {'not', 'no', 'won', 'more', 'above', 'very', 'against', 'again'}
PUNCTUATION = string.punctuation + '...'
SPECIAL_TOKEN_DICT = {"n't": 'not'}
boundary_punc = '.:;!?,'
NEG_SENT_BOUND_RE = re.compile(EMOTICON_RE.pattern + '|' + '|'.join([re.escape(punc) for punc in boundary_punc]))
NEG_WORD_RE = re.compile(r"(?:^(?:never|no|nothing|nowhere|noone|none|not)$)")
flatten = lambda l: [item for sublist in l for item in sublist]
is_word = lambda token: not(EMOTICON_RE.search(token) or token in string.punctuation or token in STOPWORDS)
class ReviewTokenizer(TreebankWordTokenizer):
_contractions = MacIntyreContractions()
CONTRACTIONS = list(map(re.compile, _contractions.CONTRACTIONS2 + _contractions.CONTRACTIONS3))
PUNCTUATION = [
(re.compile(r'([,])([^\d])'), r' \1 \2'),
(re.compile(r'([,])$'), r' \1 '),
(re.compile(r'\.\.\.'), r' ... '),
(re.compile(r'[;@#&]'), r' \g<0> '),
# Handles the final period
(re.compile(r'([^\.])(\.)([\]\)}>"\']*)\s*$'), r'\1 \2\3 '),
(re.compile(r'[?!]'), r' \g<0> '),
(re.compile(r"([^'])' "), r"\1 ' "),
]
def tokenize(self, text):
text = _replace_html_entities(text)
for regexp, substitution in self.STARTING_QUOTES:
text = regexp.sub(substitution, text)
for regexp, substitution in self.PUNCTUATION:
text = regexp.sub(substitution, text)
text = " " + text + " "
# split contractions
for regexp, substitution in self.ENDING_QUOTES:
text = regexp.sub(substitution, text)
for regexp in self.CONTRACTIONS:
text = regexp.sub(r' \1 \2 ', text)
# handle emojis
for emoticon in list(EMOTICON_RE.finditer(text))[::-1]:
pos = emoticon.span()[0]
if text[pos - 1] != ' ':
text = text[:pos] + ' ' + text[pos:]
return text.split()
def tokenize(sentence, word_tokenizer = ReviewTokenizer(), stemmer = None, lower = False, remove_punc = False, remove_stopwords = False, remove_emoji = False, convert_neg = False):
tokens = word_tokenizer.tokenize(sentence)
# convert tokens to lowercase
if lower:
tokens = [token.lower() for token in tokens]
# remove stopword tokens
if remove_stopwords:
tokens = [token for token in tokens if token not in STOPWORDS]
# remove emoji tokens
if remove_emoji:
tokens = [token for token in tokens if not EMOTICON_RE.search(token)]
# NEG_SENTENCE_BOUNDARY_LIST
if convert_neg == True:
tokens = _convert_neg(tokens)
# remove punctuation tokens
if remove_punc:
tokens = [token for token in tokens if token not in PUNCTUATION]
# stem tokens
if stemmer:
tokens = [stemmer.stem(token) for token in tokens]
tokens = [SPECIAL_TOKEN_DICT[token] if SPECIAL_TOKEN_DICT.get(token, '') else token for token in tokens]
return tokens
def _convert_neg(tokens):
'''
convert word to NEGATIVEword if there are negation words
NEG_WORD_LIST,NEG_SENTENCE_BOUNDARY_LIST
'''
new_tokenized_list = list()
i = 0
n = len(tokens)
while i < n:
token = tokens[i]
new_tokenized_list.append(token)
# e.g. if token is negative word, the following words should be negated
if NEG_WORD_RE.match(token):
i += 1
while (i < n) :
token = tokens[i]
if not NEG_SENT_BOUND_RE.match(token):
new_tokenized_list.append('neg_'+token)
else:
# end of negation
new_tokenized_list.append(token)
break
i += 1
#end while
i += 1
#end while
return new_tokenized_list
def sanitise(text):
regex_list = [
# Append whitespace after punctuations, except .com
r'(?<=[^\d]+)\{}(?=[^ \W])(?!com)',
# Reduce repeated punctuations
r'(\s*\{}\s*)+'
]
for regex_format in regex_list:
for p in '.?!':
regex = regex_format.format(p)
text = re.sub(regex, p + ' ', text)
return text
def segment_sent(text, emoji_tokenizer = TweetTokenizer()):
text = sanitise(text)
sentences = []
for sentence in sent_tokenize(text):
if EMOTICON_RE.search(sentence):
new_sentences = []
tokens = emoji_tokenizer.tokenize(sentence)
new_sentence = []
for token in tokens:
new_sentence.append(token)
if EMOTICON_RE.search(token) or token in '.?!':
new_sentences.append(' '.join(new_sentence))
new_sentence = []
if new_sentence:
new_sentences.append(' '.join(new_sentence))
sentences += new_sentences
else:
sentences.append(sentence)
if len(sentences) != 0:
if sentences[-1] in ['.', '!', '?']:
sentences[-2] = sentences[-2] + sentences[-1]
sentences = sentences[:-1]
return sentences
def plot_aspect_scores(aspect_score_df, x, y, title, x_label, y_label, showfliers=True):
sns.set(font_scale = 1.5)
fig = sns.boxplot(x=x, y=y, data=aspect_score_df, showfliers=showfliers)
plt.title(title, loc = 'center', y=1.08, fontsize = 30)
fig.set(ylim=(0, 5))
fig.set_xlabel(x_label)
fig.set_ylabel(y_label)
plt.tight_layout()
saved_path = os.path.join(IMAGES_DIRECTORY, title.lower().replace(' ', '_'))
if not os.path.exists(IMAGES_DIRECTORY):
os.makedirs(IMAGES_DIRECTORY)
fig.get_figure().savefig(saved_path, dpi=200, bbox_inches="tight")
print('{} has been saved to {}'.format(title, saved_path))
plt.close()
print('Thanks for using this aspects extraction application.')
productID = input("What is the ID of the product that you want to extract aspects from?\n")
data_path = "./data/CellPhoneReview.json"
data = []
with open(data_path) as f:
for line in f:
data.append(json.loads(line))
df = pd.DataFrame.from_dict(data)
df = df.drop(columns = ['overall', 'reviewTime', 'summary', 'unixReviewTime'])
while productID not in list(df['asin']):
print('Product {} cannot be found in our database. Please input a valid product ID, such as "B005SUHPO6", "B0042FV2SI", "B008OHNZI0" or "B00AYNRLFA"'.format(productID))
productID = input("Please enter a valid product ID.\n")
print('Start processing Aspect Extraction for product {}.'.format(productID))
print('Please wait for a while as this might take 2-3 min.')
implicit_aspect_polarity_df = pd.read_csv('implicit_aspect_polarity.csv')
aspects = list(implicit_aspect_polarity_df['aspect'].unique())
from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler()
implicit_aspect_polarity_df['polarity_intense_scaled'] = scaler.fit_transform(implicit_aspect_polarity_df[['polarity_intense']]) * 5
df = df[df['asin'] == productID]
df['sentences'] = df['reviewText'].apply(segment_sent)
df['tokenizedSentences'] = df['sentences'].apply(lambda sentences: [tokenize(sentence) for sentence in sentences])
df['tokens'] = df['tokenizedSentences'].apply(flatten)
df['words'] = df['tokens'].apply(lambda tokens: [token.lower() for token in tokens])
df['words'] = df['words'].apply(lambda tokens: [token for token in tokens if is_word(token)])
for aspect in aspects:
df[aspect] = df['words'].apply(lambda words: [word for word in words if word in list(implicit_aspect_polarity_df[implicit_aspect_polarity_df['aspect'] == aspect]['implicit'])])
product_df = df.groupby('asin')[aspects].sum().reset_index()
for aspect in aspects:
aspect_df = implicit_aspect_polarity_df[implicit_aspect_polarity_df['aspect'] == aspect]
aspect_dict = dict(zip(aspect_df['implicit'], aspect_df['polarity_intense_scaled']))
product_df[aspect + '_scores'] = product_df[aspect].apply(lambda aspect: [aspect_dict[word] for word in aspect])
aspect_list = []
aspect_score_list = []
aspect_word_list = []
for aspect in aspects:
aspects_words = product_df.iloc[0].to_dict()[aspect]
aspects_scores = product_df.iloc[0].to_dict()[aspect + '_scores']
for i in range(len(aspects_words)):
aspect_list.append(aspect)
aspect_score_list.append(aspects_scores[i])
aspect_word_list.append(aspects_words[i])
aspect_score_df = pd.DataFrame(data = {'aspect': aspect_list, 'word': aspect_word_list, 'score': aspect_score_list})
plot_aspect_scores(aspect_score_df, 'aspect', 'score', title = 'Aspect Scores for ' + productID, x_label = 'aspect', y_label = 'score', showfliers=False)