Spaces:
Sleeping
Sleeping
Upload folder using huggingface_hub
Browse files- .gitattributes +2 -0
- README.md +3 -9
- app.py +263 -0
- doc2vec_model_opinion_corpus (1).d2v +3 -0
- init.py +239 -0
- requirements.txt +0 -0
- review_detection.ipynb +1383 -0
- vercel.json +5 -0
- weights.best.from_scratch1 (1).hdf5 +3 -0
.gitattributes
CHANGED
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
36 |
+
doc2vec_model_opinion_corpus[[:space:]](1).d2v filter=lfs diff=lfs merge=lfs -text
|
37 |
+
weights.best.from_scratch1[[:space:]](1).hdf5 filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
@@ -1,12 +1,6 @@
|
|
1 |
---
|
2 |
-
title:
|
3 |
-
emoji: 👁
|
4 |
-
colorFrom: gray
|
5 |
-
colorTo: pink
|
6 |
-
sdk: gradio
|
7 |
-
sdk_version: 4.16.0
|
8 |
app_file: app.py
|
9 |
-
|
|
|
10 |
---
|
11 |
-
|
12 |
-
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
1 |
---
|
2 |
+
title: deceptive-rev
|
|
|
|
|
|
|
|
|
|
|
3 |
app_file: app.py
|
4 |
+
sdk: gradio
|
5 |
+
sdk_version: 3.44.4
|
6 |
---
|
|
|
|
app.py
ADDED
@@ -0,0 +1,263 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from flask import Flask, redirect, render_template, request, jsonify
|
2 |
+
import requests
|
3 |
+
from datetime import datetime
|
4 |
+
import pandas as pd
|
5 |
+
import numpy as np
|
6 |
+
from gensim.models import Doc2Vec
|
7 |
+
import snowballstemmer, re
|
8 |
+
from bs4 import BeautifulSoup
|
9 |
+
import re, sys
|
10 |
+
from tensorflow.keras.models import load_model
|
11 |
+
import joblib
|
12 |
+
import gradio as gr
|
13 |
+
|
14 |
+
headers = {
|
15 |
+
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36, Opera/9.80 (Windows NT 6.1; WOW64) Presto/2.12.388 Version/12.18'
|
16 |
+
}
|
17 |
+
|
18 |
+
app = Flask(__name__)
|
19 |
+
|
20 |
+
def getsoup(url):
|
21 |
+
response = requests.get(url, headers=headers)
|
22 |
+
Status_Code = response.status_code
|
23 |
+
print(url)
|
24 |
+
print(Status_Code)
|
25 |
+
|
26 |
+
if Status_Code == 200:
|
27 |
+
soup = BeautifulSoup(response.content, features="lxml")
|
28 |
+
else:
|
29 |
+
soup = getsoup(url)
|
30 |
+
return soup
|
31 |
+
|
32 |
+
def getLastPageNumber(soup, site):
|
33 |
+
pageNumber = []
|
34 |
+
if site == 'flipkart':
|
35 |
+
review_number = int(soup.find("span", "_2_R_DZ").text.strip().replace(',', '').split()[-2])
|
36 |
+
if review_number <=10:
|
37 |
+
lastPage = 1
|
38 |
+
else:
|
39 |
+
link = soup.find(attrs={"class": "_2MImiq _1Qnn1K"})
|
40 |
+
pageNumber = link.find('span').text.strip().replace(',', '').split()
|
41 |
+
lastPage1 = pageNumber[len(pageNumber)-1]
|
42 |
+
lastPage = int(lastPage1)
|
43 |
+
elif site == 'amazon':
|
44 |
+
review_number = int(soup.find("div", {"data-hook": "cr-filter-info-review-rating-count"}).text.strip().replace(',', '').split()[-3])
|
45 |
+
if review_number <=10:
|
46 |
+
lastPage = 1
|
47 |
+
else:
|
48 |
+
lastPage = review_number // 10
|
49 |
+
if lastPage > 500:
|
50 |
+
lastPage = 2
|
51 |
+
return lastPage
|
52 |
+
|
53 |
+
|
54 |
+
def geturllist(url, lastPage):
|
55 |
+
urllistPages = []
|
56 |
+
url = url[:-1]
|
57 |
+
for i in range(1,lastPage+1):
|
58 |
+
urllistPages.append (url + str(i))
|
59 |
+
return urllistPages
|
60 |
+
|
61 |
+
|
62 |
+
def getReviews(soup, site, url):
|
63 |
+
if site == 'flipkart':
|
64 |
+
#Extracting the Titles
|
65 |
+
title_sec = soup.find_all("p",'_2-N8zT')
|
66 |
+
title = []
|
67 |
+
for s in title_sec:
|
68 |
+
title.append(s.text)
|
69 |
+
|
70 |
+
author_sec = soup.find_all("p","_2sc7ZR _2V5EHH")
|
71 |
+
author = []
|
72 |
+
for r in author_sec:
|
73 |
+
author.append(r.text)
|
74 |
+
|
75 |
+
Review_text_sec = soup.find_all("div",'t-ZTKy')
|
76 |
+
text = []
|
77 |
+
for t in Review_text_sec:
|
78 |
+
text.append(t.text)
|
79 |
+
|
80 |
+
Rating = soup.find_all("div", {"class": ["_3LWZlK _1BLPMq", "_3LWZlK _32lA32 _1BLPMq", "_3LWZlK _1rdVr6 _1BLPMq"]})
|
81 |
+
rate = []
|
82 |
+
for d in Rating:
|
83 |
+
rate.append(d.text)
|
84 |
+
|
85 |
+
Date_sec = soup.find_all(lambda tag: tag.name == 'p' and tag.get('class') == ['_2sc7ZR'])
|
86 |
+
date = []
|
87 |
+
for d in Date_sec:
|
88 |
+
date.append(d.text)
|
89 |
+
|
90 |
+
help_sec = soup.find_all(lambda tag: tag.name == 'div' and tag.get('class') == ['_1LmwT9'])
|
91 |
+
help1 = []
|
92 |
+
for d in help_sec:
|
93 |
+
help1.append(d.text)
|
94 |
+
|
95 |
+
elif site == 'amazon':
|
96 |
+
n_ = 0
|
97 |
+
title_sec = soup.find_all(attrs={"data-hook": "review-title", "class": "a-size-base a-link-normal review-title a-color-base review-title-content a-text-bold"})
|
98 |
+
title = []
|
99 |
+
for s in title_sec:
|
100 |
+
title.append(s.text.replace('\n', ''))
|
101 |
+
n_ = len(title)
|
102 |
+
|
103 |
+
author_sec = soup.find_all(attrs = {"class": "a-profile-name"})
|
104 |
+
author = []
|
105 |
+
for r in author_sec:
|
106 |
+
author.append(r.text)
|
107 |
+
while(1):
|
108 |
+
if len(author) > n_:
|
109 |
+
author.pop(0)
|
110 |
+
else:
|
111 |
+
break
|
112 |
+
|
113 |
+
Review_text_sec = soup.find_all(attrs={"data-hook": "review-body", "class": "a-size-base review-text review-text-content"})
|
114 |
+
text = []
|
115 |
+
for t in Review_text_sec:
|
116 |
+
text.append(t.text.replace('\n', ''))
|
117 |
+
|
118 |
+
Rating = soup.find_all(attrs={"data-hook": "review-star-rating"})
|
119 |
+
rate = []
|
120 |
+
for d in Rating:
|
121 |
+
rate.append(d.text)
|
122 |
+
|
123 |
+
Date_sec = soup.find_all(attrs={"data-hook": "review-date"})
|
124 |
+
date = []
|
125 |
+
for d in Date_sec:
|
126 |
+
date.append(d.text)
|
127 |
+
|
128 |
+
help_sec = soup.find_all(attrs={"data-hook": "helpful-vote-statement"})
|
129 |
+
help1 = []
|
130 |
+
for d in help_sec:
|
131 |
+
help1.append(d.text.replace('\n ', ''))
|
132 |
+
while(1):
|
133 |
+
if len(help1) < n_:
|
134 |
+
help1.append(0)
|
135 |
+
else:
|
136 |
+
break
|
137 |
+
|
138 |
+
url1 = []
|
139 |
+
url1 = [url] * len(date)
|
140 |
+
|
141 |
+
collate = {'Date': date, 'URL': url1, 'Review_Title': title, 'Author': author, 'Rating': rate, 'Review_text': text, 'Review_helpful': help1}
|
142 |
+
collate_df = pd.DataFrame.from_dict(collate)
|
143 |
+
return collate_df
|
144 |
+
|
145 |
+
|
146 |
+
def preprocess_text(text):
|
147 |
+
stemmer = snowballstemmer.EnglishStemmer()
|
148 |
+
text = " ".join(stemmer.stemWords(re.sub('[!"#%\'()*+,-./:;<=>?@[\\]^_`{|}~1234567890’”“′‘\\\\]', ' ', text).split(' ')))
|
149 |
+
stop_words = set(["may", "also", "zero", "one", "two", "three", "four", "five", "six", "seven", "eight", "nine", "ten", "across","among", "beside", "however", "yet", "within"] + list('abcdefghijklmnopqrstuvwxyz'))
|
150 |
+
stop_list = stemmer.stemWords(stop_words)
|
151 |
+
stop_words.update(stop_list)
|
152 |
+
text = " ".join(filter(None, filter(lambda word: word not in stop_words, text.lower().split(' '))))
|
153 |
+
return text.split(' ')
|
154 |
+
|
155 |
+
def vectorize_comments_(df, d2v_model):
|
156 |
+
y = []
|
157 |
+
comments = []
|
158 |
+
for i in range(0, len(df)):
|
159 |
+
|
160 |
+
print(i)
|
161 |
+
label = 'SENT_%s' %i
|
162 |
+
comments.append(d2v_model.docvecs[label])
|
163 |
+
return comments
|
164 |
+
|
165 |
+
def scraper(url):
|
166 |
+
df2 = []
|
167 |
+
soup = getsoup(url)
|
168 |
+
site = url.split('.')[1]
|
169 |
+
if site == 'flipkart':
|
170 |
+
url = url + '&page=1'
|
171 |
+
elif site == 'amazon':
|
172 |
+
url = url + '&pageNumber=1'
|
173 |
+
product = url.split('/')[3]
|
174 |
+
lastPage = 1
|
175 |
+
urllistPages = geturllist(url, lastPage)
|
176 |
+
x = 1
|
177 |
+
for url in urllistPages:
|
178 |
+
soup = getsoup(url)
|
179 |
+
df1 = getReviews(soup, site, url)
|
180 |
+
if x == 1:
|
181 |
+
df3 = []
|
182 |
+
df3 = df1
|
183 |
+
else:
|
184 |
+
df2 = df3
|
185 |
+
result = df2.append(df1, ignore_index=True)
|
186 |
+
df3 = result
|
187 |
+
x += 1
|
188 |
+
|
189 |
+
loaded_model = load_model('weights.best.from_scratch1 (1).hdf5')
|
190 |
+
|
191 |
+
preprocessed_arr = [preprocess_text(x) for x in list(df3['Review_text'])]
|
192 |
+
|
193 |
+
doc2vec_model = Doc2Vec.load("doc2vec_model_opinion_corpus (1).d2v")
|
194 |
+
|
195 |
+
textData = vectorize_comments_(preprocessed_arr, doc2vec_model)
|
196 |
+
|
197 |
+
textData_array = np.array(textData)
|
198 |
+
|
199 |
+
num_vectors = textData_array.shape[0]
|
200 |
+
textData_3d = textData_array.reshape((num_vectors, 1, -1))
|
201 |
+
|
202 |
+
new_shape = (textData_array.shape[0], 380, 512)
|
203 |
+
|
204 |
+
X_test3_reshaped = np.zeros(new_shape, dtype=textData_3d.dtype)
|
205 |
+
X_test3_reshaped[:, :textData_3d.shape[1], :textData_3d.shape[2]] = textData_3d
|
206 |
+
|
207 |
+
predictions = np.rint(loaded_model.predict(X_test3_reshaped))
|
208 |
+
|
209 |
+
argMax = []
|
210 |
+
|
211 |
+
for i in predictions:
|
212 |
+
argMax.append(np.argmax(i))
|
213 |
+
|
214 |
+
print(argMax)
|
215 |
+
print(list(df3['Review_text'])[3])
|
216 |
+
|
217 |
+
arr = []
|
218 |
+
for i, j in enumerate(argMax):
|
219 |
+
if j == 2 or j == 1:
|
220 |
+
arr.append(list(df3['Review_text'])[i])
|
221 |
+
return arr
|
222 |
+
|
223 |
+
|
224 |
+
# @app.route('/', methods=['GET'])
|
225 |
+
# def index():
|
226 |
+
# results = []
|
227 |
+
# if request.args.get('url'):
|
228 |
+
# results = scraper(request.args.get('url'))
|
229 |
+
# return results
|
230 |
+
|
231 |
+
|
232 |
+
# if __name__ == "__main__":
|
233 |
+
# app.run(debug=True)
|
234 |
+
|
235 |
+
|
236 |
+
|
237 |
+
def index():
|
238 |
+
results = []
|
239 |
+
if request.args.get('url'):
|
240 |
+
results = scraper(request.args.get('url'))
|
241 |
+
return results
|
242 |
+
|
243 |
+
|
244 |
+
inputs_image_url = [
|
245 |
+
gr.Textbox(type="text", label="Image URL"),
|
246 |
+
]
|
247 |
+
|
248 |
+
outputs_result_dict = [
|
249 |
+
gr.Textbox(type="text", label="Result Dictionary"),
|
250 |
+
]
|
251 |
+
|
252 |
+
interface_image_url = gr.Interface(
|
253 |
+
fn=index,
|
254 |
+
inputs=inputs_image_url,
|
255 |
+
outputs=outputs_result_dict,
|
256 |
+
title="Dark review detection",
|
257 |
+
cache_examples=False,
|
258 |
+
)
|
259 |
+
|
260 |
+
gr.TabbedInterface(
|
261 |
+
[interface_image_url],
|
262 |
+
tab_names=['Reviews inference']
|
263 |
+
).queue().launch()
|
doc2vec_model_opinion_corpus (1).d2v
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:deabad8d2bf4677f8f6f7069da1f928f6ce1fe45722f0839c99a2615f37f28ea
|
3 |
+
size 29196813
|
init.py
ADDED
@@ -0,0 +1,239 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import pandas as pd
|
2 |
+
import numpy as np
|
3 |
+
from string import ascii_lowercase
|
4 |
+
from gensim.models import Doc2Vec
|
5 |
+
from gensim.models import doc2vec
|
6 |
+
from gensim.models import KeyedVectors
|
7 |
+
import snowballstemmer, re
|
8 |
+
import requests
|
9 |
+
from bs4 import BeautifulSoup
|
10 |
+
import re, sys
|
11 |
+
from tensorflow.keras.models import load_model
|
12 |
+
import joblib
|
13 |
+
|
14 |
+
headers = {
|
15 |
+
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36, Opera/9.80 (Windows NT 6.1; WOW64) Presto/2.12.388 Version/12.18'
|
16 |
+
}
|
17 |
+
|
18 |
+
def getsoup(url):
|
19 |
+
response = requests.get(url, headers=headers)
|
20 |
+
Status_Code = response.status_code
|
21 |
+
print(url)
|
22 |
+
print(Status_Code)
|
23 |
+
|
24 |
+
if Status_Code == 200:
|
25 |
+
soup = BeautifulSoup(response.content, features="lxml")
|
26 |
+
else:
|
27 |
+
soup = getsoup(url)
|
28 |
+
return soup
|
29 |
+
|
30 |
+
#Get Last Page number
|
31 |
+
def getLastPageNumber(soup, site):
|
32 |
+
pageNumber = []
|
33 |
+
if site == 'flipkart':
|
34 |
+
review_number = int(soup.find("span", "_2_R_DZ").text.strip().replace(',', '').split()[-2])
|
35 |
+
if review_number <=10:
|
36 |
+
lastPage = 1
|
37 |
+
else:
|
38 |
+
link = soup.find(attrs={"class": "_2MImiq _1Qnn1K"})
|
39 |
+
pageNumber = link.find('span').text.strip().replace(',', '').split()
|
40 |
+
lastPage1 = pageNumber[len(pageNumber)-1]
|
41 |
+
lastPage = int(lastPage1)
|
42 |
+
elif site == 'amazon':
|
43 |
+
review_number = int(soup.find("div", {"data-hook": "cr-filter-info-review-rating-count"}).text.strip().replace(',', '').split()[-3])
|
44 |
+
if review_number <=10:
|
45 |
+
lastPage = 1
|
46 |
+
else:
|
47 |
+
lastPage = review_number // 10
|
48 |
+
if lastPage > 500:
|
49 |
+
lastPage = 2
|
50 |
+
return lastPage
|
51 |
+
|
52 |
+
|
53 |
+
def geturllist(url, lastPage):
|
54 |
+
urllistPages = []
|
55 |
+
url = url[:-1]
|
56 |
+
for i in range(1,lastPage+1):
|
57 |
+
urllistPages.append (url + str(i))
|
58 |
+
return urllistPages
|
59 |
+
|
60 |
+
|
61 |
+
def getReviews(soup, site, url):
|
62 |
+
if site == 'flipkart':
|
63 |
+
#Extracting the Titles
|
64 |
+
title_sec = soup.find_all("p",'_2-N8zT')
|
65 |
+
title = []
|
66 |
+
for s in title_sec:
|
67 |
+
title.append(s.text)
|
68 |
+
|
69 |
+
#Extracting the Author names
|
70 |
+
author_sec = soup.find_all("p","_2sc7ZR _2V5EHH")
|
71 |
+
author = []
|
72 |
+
for r in author_sec:
|
73 |
+
author.append(r.text)
|
74 |
+
|
75 |
+
#Extracting the Text
|
76 |
+
Review_text_sec = soup.find_all("div",'t-ZTKy')
|
77 |
+
text = []
|
78 |
+
for t in Review_text_sec:
|
79 |
+
text.append(t.text)
|
80 |
+
|
81 |
+
#Extracting the Star rating
|
82 |
+
Rating = soup.find_all("div", {"class": ["_3LWZlK _1BLPMq", "_3LWZlK _32lA32 _1BLPMq", "_3LWZlK _1rdVr6 _1BLPMq"]})
|
83 |
+
rate = []
|
84 |
+
for d in Rating:
|
85 |
+
rate.append(d.text)
|
86 |
+
|
87 |
+
#Extracting the Date
|
88 |
+
Date_sec = soup.find_all(lambda tag: tag.name == 'p' and tag.get('class') == ['_2sc7ZR'])
|
89 |
+
date = []
|
90 |
+
for d in Date_sec:
|
91 |
+
date.append(d.text)
|
92 |
+
|
93 |
+
#Extracting the Helpful rating
|
94 |
+
help_sec = soup.find_all(lambda tag: tag.name == 'div' and tag.get('class') == ['_1LmwT9'])
|
95 |
+
help1 = []
|
96 |
+
for d in help_sec:
|
97 |
+
help1.append(d.text)
|
98 |
+
|
99 |
+
elif site == 'amazon':
|
100 |
+
n_ = 0
|
101 |
+
title_sec = soup.find_all(attrs={"data-hook": "review-title", "class": "a-size-base a-link-normal review-title a-color-base review-title-content a-text-bold"})
|
102 |
+
title = []
|
103 |
+
for s in title_sec:
|
104 |
+
title.append(s.text.replace('\n', ''))
|
105 |
+
n_ = len(title)
|
106 |
+
|
107 |
+
author_sec = soup.find_all(attrs = {"class": "a-profile-name"})
|
108 |
+
author = []
|
109 |
+
for r in author_sec:
|
110 |
+
author.append(r.text)
|
111 |
+
while(1):
|
112 |
+
if len(author) > n_:
|
113 |
+
author.pop(0)
|
114 |
+
else:
|
115 |
+
break
|
116 |
+
|
117 |
+
Review_text_sec = soup.find_all(attrs={"data-hook": "review-body", "class": "a-size-base review-text review-text-content"})
|
118 |
+
text = []
|
119 |
+
for t in Review_text_sec:
|
120 |
+
text.append(t.text.replace('\n', ''))
|
121 |
+
|
122 |
+
Rating = soup.find_all(attrs={"data-hook": "review-star-rating"})
|
123 |
+
rate = []
|
124 |
+
for d in Rating:
|
125 |
+
rate.append(d.text)
|
126 |
+
|
127 |
+
Date_sec = soup.find_all(attrs={"data-hook": "review-date"})
|
128 |
+
date = []
|
129 |
+
for d in Date_sec:
|
130 |
+
date.append(d.text)
|
131 |
+
|
132 |
+
help_sec = soup.find_all(attrs={"data-hook": "helpful-vote-statement"})
|
133 |
+
help1 = []
|
134 |
+
for d in help_sec:
|
135 |
+
help1.append(d.text.replace('\n ', ''))
|
136 |
+
while(1):
|
137 |
+
if len(help1) < n_:
|
138 |
+
help1.append(0)
|
139 |
+
else:
|
140 |
+
break
|
141 |
+
|
142 |
+
url1 = []
|
143 |
+
url1 = [url] * len(date)
|
144 |
+
|
145 |
+
collate = {'Date': date, 'URL': url1, 'Review_Title': title, 'Author': author, 'Rating': rate, 'Review_text': text, 'Review_helpful': help1}
|
146 |
+
collate_df = pd.DataFrame.from_dict(collate)
|
147 |
+
return collate_df
|
148 |
+
|
149 |
+
def scraper(url):
|
150 |
+
df2 = []
|
151 |
+
soup = getsoup(url)
|
152 |
+
site = url.split('.')[1]
|
153 |
+
if site == 'flipkart':
|
154 |
+
url = url + '&page=1'
|
155 |
+
elif site == 'amazon':
|
156 |
+
url = url + '&pageNumber=1'
|
157 |
+
product = url.split('/')[3]
|
158 |
+
lastPage = 1
|
159 |
+
urllistPages = geturllist(url, lastPage)
|
160 |
+
x = 1
|
161 |
+
for url in urllistPages:
|
162 |
+
soup = getsoup(url)
|
163 |
+
df1 = getReviews(soup, site, url)
|
164 |
+
if x == 1:
|
165 |
+
df3 = []
|
166 |
+
df3 = df1
|
167 |
+
else:
|
168 |
+
df2 = df3
|
169 |
+
result = df2.append(df1, ignore_index=True)
|
170 |
+
df3 = result
|
171 |
+
x += 1
|
172 |
+
print(list(df3['Review_text']))
|
173 |
+
return list(df3['Review_text'])
|
174 |
+
|
175 |
+
arr = scraper('https://www.amazon.in/Redmi-inches-Ready-Smart-L32R8-FVIN/product-review/B0BVMLNGXR/ref=lp_90117314031_1_1?pf_rd_p=9e034799-55e2-4ab2-b0d0-eb42f95b2d05&pf_rd_r=V81TJ2VTRM0BYHQ6XX8S&sbo=RZvfv%2F%2FHxDF%2BO5021pAnSA%3D%3D&th=1')
|
176 |
+
|
177 |
+
TaggedDocument = doc2vec.TaggedDocument
|
178 |
+
|
179 |
+
loaded_model = load_model('weights.best.from_scratch1 (1).hdf5')
|
180 |
+
|
181 |
+
def preprocess_text(text):
|
182 |
+
stemmer = snowballstemmer.EnglishStemmer()
|
183 |
+
text = " ".join(stemmer.stemWords(re.sub('[!"#%\'()*+,-./:;<=>?@[\\]^_`{|}~1234567890’”“′‘\\\\]', ' ', text).split(' ')))
|
184 |
+
|
185 |
+
stop_words = set(["may", "also", "zero", "one", "two", "three", "four", "five", "six", "seven", "eight", "nine", "ten", "across", "among", "beside", "however", "yet", "within"] + list(ascii_lowercase))
|
186 |
+
stop_list = stemmer.stemWords(stop_words)
|
187 |
+
stop_words.update(stop_list)
|
188 |
+
text = " ".join(filter(None, filter(lambda word: word not in stop_words, text.lower().split(' '))))
|
189 |
+
|
190 |
+
return text.split(' ')
|
191 |
+
|
192 |
+
|
193 |
+
# arr = ['This is not a nice product', 'This is brilliant product', "I have had this tv for a year and a half. The tv worked smoothly and recently, it's display stopped working. I have contacted the company and they assured a repair within 8-10 days replacing the display panel. Then came the worst after sale experience I had ever. They didn't respond to my calls and whenever, I raised a grievance, I was just given different dates and extensions.", ''' In monotheistic belief systems, God is usually viewed as the supreme being, creator, and principal object of faith. In polytheistic belief systems, a god is "a spirit or being believed to have created, or for controlling some part of the universe or life, for which such a deity is often worshipped ''']
|
194 |
+
|
195 |
+
preprocessed_arr = [preprocess_text(x) for x in arr]
|
196 |
+
|
197 |
+
doc2vec_model = Doc2Vec.load("doc2vec_model_opinion_corpus (1).d2v")
|
198 |
+
|
199 |
+
def vectorize_comments_(df, d2v_model):
|
200 |
+
y = []
|
201 |
+
comments = []
|
202 |
+
for i in range(0, len(df)):
|
203 |
+
print(i)
|
204 |
+
label = 'SENT_%s' %i
|
205 |
+
comments.append(d2v_model.docvecs[label])
|
206 |
+
|
207 |
+
return comments
|
208 |
+
|
209 |
+
textData = vectorize_comments_(preprocessed_arr, doc2vec_model)
|
210 |
+
|
211 |
+
import numpy as np
|
212 |
+
|
213 |
+
textData_array = np.array(textData)
|
214 |
+
|
215 |
+
num_vectors = textData_array.shape[0]
|
216 |
+
textData_3d = textData_array.reshape((num_vectors, 1, -1))
|
217 |
+
|
218 |
+
new_shape = (textData_array.shape[0], 380, 512)
|
219 |
+
|
220 |
+
X_test3_reshaped = np.zeros(new_shape, dtype=textData_3d.dtype)
|
221 |
+
X_test3_reshaped[:, :textData_3d.shape[1], :textData_3d.shape[2]] = textData_3d
|
222 |
+
|
223 |
+
predictions = np.rint(loaded_model.predict(X_test3_reshaped))
|
224 |
+
|
225 |
+
argMax = []
|
226 |
+
|
227 |
+
for i in predictions:
|
228 |
+
argMax.append(np.argmax(i))
|
229 |
+
|
230 |
+
def returnRequirements(comments, preds):
|
231 |
+
arr = []
|
232 |
+
for i, j in enumerate(preds):
|
233 |
+
if j == 3 or j == 0:
|
234 |
+
arr.append(comments[i])
|
235 |
+
return arr
|
236 |
+
|
237 |
+
# 3 & 0 is deceptive rest aren't
|
238 |
+
print(argMax)
|
239 |
+
print(returnRequirements(arr, argMax))
|
requirements.txt
ADDED
Binary file (790 Bytes). View file
|
|
review_detection.ipynb
ADDED
@@ -0,0 +1,1383 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "code",
|
5 |
+
"execution_count": null,
|
6 |
+
"metadata": {
|
7 |
+
"colab": {
|
8 |
+
"base_uri": "https://localhost:8080/"
|
9 |
+
},
|
10 |
+
"id": "aqGqpcYIpf_q",
|
11 |
+
"outputId": "a6d87acc-4df9-4abf-973c-c791c5461af9"
|
12 |
+
},
|
13 |
+
"outputs": [
|
14 |
+
{
|
15 |
+
"name": "stdout",
|
16 |
+
"output_type": "stream",
|
17 |
+
"text": [
|
18 |
+
"Downloading deceptive-opinion-spam-corpus.zip to /content\n",
|
19 |
+
"\r 0% 0.00/456k [00:00<?, ?B/s]\n",
|
20 |
+
"\r100% 456k/456k [00:00<00:00, 111MB/s]\n"
|
21 |
+
]
|
22 |
+
}
|
23 |
+
],
|
24 |
+
"source": [
|
25 |
+
"!kaggle datasets download -d rtatman/deceptive-opinion-spam-corpus"
|
26 |
+
]
|
27 |
+
},
|
28 |
+
{
|
29 |
+
"cell_type": "code",
|
30 |
+
"execution_count": null,
|
31 |
+
"metadata": {
|
32 |
+
"id": "zu4NTMHapmms"
|
33 |
+
},
|
34 |
+
"outputs": [],
|
35 |
+
"source": [
|
36 |
+
"import zipfile\n",
|
37 |
+
"zip_ref = zipfile.ZipFile('/content/deceptive-opinion-spam-corpus.zip', 'r')\n",
|
38 |
+
"zip_ref.extractall('/content')\n",
|
39 |
+
"zip_ref.close()"
|
40 |
+
]
|
41 |
+
},
|
42 |
+
{
|
43 |
+
"cell_type": "code",
|
44 |
+
"execution_count": null,
|
45 |
+
"metadata": {
|
46 |
+
"id": "3hIl0gHep9q6"
|
47 |
+
},
|
48 |
+
"outputs": [],
|
49 |
+
"source": [
|
50 |
+
"import pandas as pd\n",
|
51 |
+
"import numpy as np\n",
|
52 |
+
"from keras.preprocessing import sequence\n",
|
53 |
+
"from keras.layers import TimeDistributed, GlobalAveragePooling1D, GlobalAveragePooling2D, BatchNormalization\n",
|
54 |
+
"from keras.layers import LSTM\n",
|
55 |
+
"from keras.layers import Conv1D, MaxPooling1D, Conv2D, MaxPooling2D, AveragePooling1D\n",
|
56 |
+
"from keras.layers import Embedding\n",
|
57 |
+
"from keras.layers import Dropout, Flatten, Bidirectional, Dense, Activation, TimeDistributed\n",
|
58 |
+
"from keras.models import Model, Sequential\n",
|
59 |
+
"from tensorflow.keras.utils import to_categorical\n",
|
60 |
+
"from sklearn.model_selection import train_test_split\n",
|
61 |
+
"from sklearn.preprocessing import LabelEncoder\n",
|
62 |
+
"from nltk.corpus import stopwords\n",
|
63 |
+
"from nltk.tokenize import word_tokenize, sent_tokenize\n",
|
64 |
+
"from nltk.stem.wordnet import WordNetLemmatizer\n",
|
65 |
+
"from string import ascii_lowercase\n",
|
66 |
+
"from collections import Counter\n",
|
67 |
+
"from gensim.models import Word2Vec\n",
|
68 |
+
"from gensim.models import Doc2Vec\n",
|
69 |
+
"from gensim.models import doc2vec\n",
|
70 |
+
"from gensim.models import KeyedVectors\n",
|
71 |
+
"import itertools, nltk, snowballstemmer, re\n",
|
72 |
+
"import random\n",
|
73 |
+
"\n",
|
74 |
+
"TaggedDocument = doc2vec.TaggedDocument"
|
75 |
+
]
|
76 |
+
},
|
77 |
+
{
|
78 |
+
"cell_type": "code",
|
79 |
+
"execution_count": null,
|
80 |
+
"metadata": {
|
81 |
+
"id": "fzwpJb7EqCjc"
|
82 |
+
},
|
83 |
+
"outputs": [],
|
84 |
+
"source": [
|
85 |
+
"class LabeledLineSentence(object):\n",
|
86 |
+
" def __init__(self, sources):\n",
|
87 |
+
" self.sources = sources\n",
|
88 |
+
"\n",
|
89 |
+
" flipped = {}\n",
|
90 |
+
"\n",
|
91 |
+
" for key, value in sources.items():\n",
|
92 |
+
" if value not in flipped:\n",
|
93 |
+
" flipped[value] = [key]\n",
|
94 |
+
" else:\n",
|
95 |
+
" raise Exception('Non-unique prefix encountered')\n",
|
96 |
+
"\n",
|
97 |
+
" def __iter__(self):\n",
|
98 |
+
" for source, prefix in self.sources.items():\n",
|
99 |
+
" with utils.smart_open(source) as fin:\n",
|
100 |
+
" for item_no, line in enumerate(fin):\n",
|
101 |
+
" yield TaggedDocument(utils.to_unicode(line).split(), [prefix + '_%s' % item_no])\n",
|
102 |
+
"\n",
|
103 |
+
" def to_array(self):\n",
|
104 |
+
" self.sentences = []\n",
|
105 |
+
" for source, prefix in self.sources.items():\n",
|
106 |
+
" with utils.smart_open(source) as fin:\n",
|
107 |
+
" for item_no, line in enumerate(fin):\n",
|
108 |
+
" self.sentences.append(TaggedDocument(utils.to_unicode(line).split(), [prefix + '_%s' % item_no]))\n",
|
109 |
+
" return self.sentences\n",
|
110 |
+
"\n",
|
111 |
+
" def sentences_perm(self):\n",
|
112 |
+
" shuffled = list(self.sentences)\n",
|
113 |
+
" random.shuffle(shuffled)\n",
|
114 |
+
" return shuffled"
|
115 |
+
]
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"cell_type": "code",
|
119 |
+
"execution_count": null,
|
120 |
+
"metadata": {
|
121 |
+
"id": "e_auMclPqmrI"
|
122 |
+
},
|
123 |
+
"outputs": [],
|
124 |
+
"source": [
|
125 |
+
"data = pd.read_csv(\"/content/deceptive-opinion.csv\")"
|
126 |
+
]
|
127 |
+
},
|
128 |
+
{
|
129 |
+
"cell_type": "code",
|
130 |
+
"execution_count": null,
|
131 |
+
"metadata": {
|
132 |
+
"id": "G-EIqcKVrbEl"
|
133 |
+
},
|
134 |
+
"outputs": [],
|
135 |
+
"source": [
|
136 |
+
"data['polarity'] = np.where(data['polarity']=='positive', 1, 0)\n",
|
137 |
+
"data['deceptive'] = np.where(data['deceptive']=='truthful', 1, 0)"
|
138 |
+
]
|
139 |
+
},
|
140 |
+
{
|
141 |
+
"cell_type": "code",
|
142 |
+
"execution_count": null,
|
143 |
+
"metadata": {
|
144 |
+
"colab": {
|
145 |
+
"base_uri": "https://localhost:8080/",
|
146 |
+
"height": 300
|
147 |
+
},
|
148 |
+
"id": "2REEjGj9rck1",
|
149 |
+
"outputId": "a5679d0f-f0b9-4c21-8005-42058e2cc4fc"
|
150 |
+
},
|
151 |
+
"outputs": [
|
152 |
+
{
|
153 |
+
"data": {
|
154 |
+
"text/html": [
|
155 |
+
"\n",
|
156 |
+
" <div id=\"df-6614128d-b81e-41f1-9a92-25467585644c\" class=\"colab-df-container\">\n",
|
157 |
+
" <div>\n",
|
158 |
+
"<style scoped>\n",
|
159 |
+
" .dataframe tbody tr th:only-of-type {\n",
|
160 |
+
" vertical-align: middle;\n",
|
161 |
+
" }\n",
|
162 |
+
"\n",
|
163 |
+
" .dataframe tbody tr th {\n",
|
164 |
+
" vertical-align: top;\n",
|
165 |
+
" }\n",
|
166 |
+
"\n",
|
167 |
+
" .dataframe thead th {\n",
|
168 |
+
" text-align: right;\n",
|
169 |
+
" }\n",
|
170 |
+
"</style>\n",
|
171 |
+
"<table border=\"1\" class=\"dataframe\">\n",
|
172 |
+
" <thead>\n",
|
173 |
+
" <tr style=\"text-align: right;\">\n",
|
174 |
+
" <th></th>\n",
|
175 |
+
" <th>deceptive</th>\n",
|
176 |
+
" <th>polarity</th>\n",
|
177 |
+
" </tr>\n",
|
178 |
+
" </thead>\n",
|
179 |
+
" <tbody>\n",
|
180 |
+
" <tr>\n",
|
181 |
+
" <th>count</th>\n",
|
182 |
+
" <td>1600.000000</td>\n",
|
183 |
+
" <td>1600.000000</td>\n",
|
184 |
+
" </tr>\n",
|
185 |
+
" <tr>\n",
|
186 |
+
" <th>mean</th>\n",
|
187 |
+
" <td>0.500000</td>\n",
|
188 |
+
" <td>0.500000</td>\n",
|
189 |
+
" </tr>\n",
|
190 |
+
" <tr>\n",
|
191 |
+
" <th>std</th>\n",
|
192 |
+
" <td>0.500156</td>\n",
|
193 |
+
" <td>0.500156</td>\n",
|
194 |
+
" </tr>\n",
|
195 |
+
" <tr>\n",
|
196 |
+
" <th>min</th>\n",
|
197 |
+
" <td>0.000000</td>\n",
|
198 |
+
" <td>0.000000</td>\n",
|
199 |
+
" </tr>\n",
|
200 |
+
" <tr>\n",
|
201 |
+
" <th>25%</th>\n",
|
202 |
+
" <td>0.000000</td>\n",
|
203 |
+
" <td>0.000000</td>\n",
|
204 |
+
" </tr>\n",
|
205 |
+
" <tr>\n",
|
206 |
+
" <th>50%</th>\n",
|
207 |
+
" <td>0.500000</td>\n",
|
208 |
+
" <td>0.500000</td>\n",
|
209 |
+
" </tr>\n",
|
210 |
+
" <tr>\n",
|
211 |
+
" <th>75%</th>\n",
|
212 |
+
" <td>1.000000</td>\n",
|
213 |
+
" <td>1.000000</td>\n",
|
214 |
+
" </tr>\n",
|
215 |
+
" <tr>\n",
|
216 |
+
" <th>max</th>\n",
|
217 |
+
" <td>1.000000</td>\n",
|
218 |
+
" <td>1.000000</td>\n",
|
219 |
+
" </tr>\n",
|
220 |
+
" </tbody>\n",
|
221 |
+
"</table>\n",
|
222 |
+
"</div>\n",
|
223 |
+
" <div class=\"colab-df-buttons\">\n",
|
224 |
+
"\n",
|
225 |
+
" <div class=\"colab-df-container\">\n",
|
226 |
+
" <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-6614128d-b81e-41f1-9a92-25467585644c')\"\n",
|
227 |
+
" title=\"Convert this dataframe to an interactive table.\"\n",
|
228 |
+
" style=\"display:none;\">\n",
|
229 |
+
"\n",
|
230 |
+
" <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
|
231 |
+
" <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
|
232 |
+
" </svg>\n",
|
233 |
+
" </button>\n",
|
234 |
+
"\n",
|
235 |
+
" <style>\n",
|
236 |
+
" .colab-df-container {\n",
|
237 |
+
" display:flex;\n",
|
238 |
+
" gap: 12px;\n",
|
239 |
+
" }\n",
|
240 |
+
"\n",
|
241 |
+
" .colab-df-convert {\n",
|
242 |
+
" background-color: #E8F0FE;\n",
|
243 |
+
" border: none;\n",
|
244 |
+
" border-radius: 50%;\n",
|
245 |
+
" cursor: pointer;\n",
|
246 |
+
" display: none;\n",
|
247 |
+
" fill: #1967D2;\n",
|
248 |
+
" height: 32px;\n",
|
249 |
+
" padding: 0 0 0 0;\n",
|
250 |
+
" width: 32px;\n",
|
251 |
+
" }\n",
|
252 |
+
"\n",
|
253 |
+
" .colab-df-convert:hover {\n",
|
254 |
+
" background-color: #E2EBFA;\n",
|
255 |
+
" box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
|
256 |
+
" fill: #174EA6;\n",
|
257 |
+
" }\n",
|
258 |
+
"\n",
|
259 |
+
" .colab-df-buttons div {\n",
|
260 |
+
" margin-bottom: 4px;\n",
|
261 |
+
" }\n",
|
262 |
+
"\n",
|
263 |
+
" [theme=dark] .colab-df-convert {\n",
|
264 |
+
" background-color: #3B4455;\n",
|
265 |
+
" fill: #D2E3FC;\n",
|
266 |
+
" }\n",
|
267 |
+
"\n",
|
268 |
+
" [theme=dark] .colab-df-convert:hover {\n",
|
269 |
+
" background-color: #434B5C;\n",
|
270 |
+
" box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
|
271 |
+
" filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
|
272 |
+
" fill: #FFFFFF;\n",
|
273 |
+
" }\n",
|
274 |
+
" </style>\n",
|
275 |
+
"\n",
|
276 |
+
" <script>\n",
|
277 |
+
" const buttonEl =\n",
|
278 |
+
" document.querySelector('#df-6614128d-b81e-41f1-9a92-25467585644c button.colab-df-convert');\n",
|
279 |
+
" buttonEl.style.display =\n",
|
280 |
+
" google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
|
281 |
+
"\n",
|
282 |
+
" async function convertToInteractive(key) {\n",
|
283 |
+
" const element = document.querySelector('#df-6614128d-b81e-41f1-9a92-25467585644c');\n",
|
284 |
+
" const dataTable =\n",
|
285 |
+
" await google.colab.kernel.invokeFunction('convertToInteractive',\n",
|
286 |
+
" [key], {});\n",
|
287 |
+
" if (!dataTable) return;\n",
|
288 |
+
"\n",
|
289 |
+
" const docLinkHtml = 'Like what you see? Visit the ' +\n",
|
290 |
+
" '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
|
291 |
+
" + ' to learn more about interactive tables.';\n",
|
292 |
+
" element.innerHTML = '';\n",
|
293 |
+
" dataTable['output_type'] = 'display_data';\n",
|
294 |
+
" await google.colab.output.renderOutput(dataTable, element);\n",
|
295 |
+
" const docLink = document.createElement('div');\n",
|
296 |
+
" docLink.innerHTML = docLinkHtml;\n",
|
297 |
+
" element.appendChild(docLink);\n",
|
298 |
+
" }\n",
|
299 |
+
" </script>\n",
|
300 |
+
" </div>\n",
|
301 |
+
"\n",
|
302 |
+
"\n",
|
303 |
+
"<div id=\"df-cc66d435-72a8-4146-ba22-ea8ef5610445\">\n",
|
304 |
+
" <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-cc66d435-72a8-4146-ba22-ea8ef5610445')\"\n",
|
305 |
+
" title=\"Suggest charts\"\n",
|
306 |
+
" style=\"display:none;\">\n",
|
307 |
+
"\n",
|
308 |
+
"<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
|
309 |
+
" width=\"24px\">\n",
|
310 |
+
" <g>\n",
|
311 |
+
" <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
|
312 |
+
" </g>\n",
|
313 |
+
"</svg>\n",
|
314 |
+
" </button>\n",
|
315 |
+
"\n",
|
316 |
+
"<style>\n",
|
317 |
+
" .colab-df-quickchart {\n",
|
318 |
+
" --bg-color: #E8F0FE;\n",
|
319 |
+
" --fill-color: #1967D2;\n",
|
320 |
+
" --hover-bg-color: #E2EBFA;\n",
|
321 |
+
" --hover-fill-color: #174EA6;\n",
|
322 |
+
" --disabled-fill-color: #AAA;\n",
|
323 |
+
" --disabled-bg-color: #DDD;\n",
|
324 |
+
" }\n",
|
325 |
+
"\n",
|
326 |
+
" [theme=dark] .colab-df-quickchart {\n",
|
327 |
+
" --bg-color: #3B4455;\n",
|
328 |
+
" --fill-color: #D2E3FC;\n",
|
329 |
+
" --hover-bg-color: #434B5C;\n",
|
330 |
+
" --hover-fill-color: #FFFFFF;\n",
|
331 |
+
" --disabled-bg-color: #3B4455;\n",
|
332 |
+
" --disabled-fill-color: #666;\n",
|
333 |
+
" }\n",
|
334 |
+
"\n",
|
335 |
+
" .colab-df-quickchart {\n",
|
336 |
+
" background-color: var(--bg-color);\n",
|
337 |
+
" border: none;\n",
|
338 |
+
" border-radius: 50%;\n",
|
339 |
+
" cursor: pointer;\n",
|
340 |
+
" display: none;\n",
|
341 |
+
" fill: var(--fill-color);\n",
|
342 |
+
" height: 32px;\n",
|
343 |
+
" padding: 0;\n",
|
344 |
+
" width: 32px;\n",
|
345 |
+
" }\n",
|
346 |
+
"\n",
|
347 |
+
" .colab-df-quickchart:hover {\n",
|
348 |
+
" background-color: var(--hover-bg-color);\n",
|
349 |
+
" box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
|
350 |
+
" fill: var(--button-hover-fill-color);\n",
|
351 |
+
" }\n",
|
352 |
+
"\n",
|
353 |
+
" .colab-df-quickchart-complete:disabled,\n",
|
354 |
+
" .colab-df-quickchart-complete:disabled:hover {\n",
|
355 |
+
" background-color: var(--disabled-bg-color);\n",
|
356 |
+
" fill: var(--disabled-fill-color);\n",
|
357 |
+
" box-shadow: none;\n",
|
358 |
+
" }\n",
|
359 |
+
"\n",
|
360 |
+
" .colab-df-spinner {\n",
|
361 |
+
" border: 2px solid var(--fill-color);\n",
|
362 |
+
" border-color: transparent;\n",
|
363 |
+
" border-bottom-color: var(--fill-color);\n",
|
364 |
+
" animation:\n",
|
365 |
+
" spin 1s steps(1) infinite;\n",
|
366 |
+
" }\n",
|
367 |
+
"\n",
|
368 |
+
" @keyframes spin {\n",
|
369 |
+
" 0% {\n",
|
370 |
+
" border-color: transparent;\n",
|
371 |
+
" border-bottom-color: var(--fill-color);\n",
|
372 |
+
" border-left-color: var(--fill-color);\n",
|
373 |
+
" }\n",
|
374 |
+
" 20% {\n",
|
375 |
+
" border-color: transparent;\n",
|
376 |
+
" border-left-color: var(--fill-color);\n",
|
377 |
+
" border-top-color: var(--fill-color);\n",
|
378 |
+
" }\n",
|
379 |
+
" 30% {\n",
|
380 |
+
" border-color: transparent;\n",
|
381 |
+
" border-left-color: var(--fill-color);\n",
|
382 |
+
" border-top-color: var(--fill-color);\n",
|
383 |
+
" border-right-color: var(--fill-color);\n",
|
384 |
+
" }\n",
|
385 |
+
" 40% {\n",
|
386 |
+
" border-color: transparent;\n",
|
387 |
+
" border-right-color: var(--fill-color);\n",
|
388 |
+
" border-top-color: var(--fill-color);\n",
|
389 |
+
" }\n",
|
390 |
+
" 60% {\n",
|
391 |
+
" border-color: transparent;\n",
|
392 |
+
" border-right-color: var(--fill-color);\n",
|
393 |
+
" }\n",
|
394 |
+
" 80% {\n",
|
395 |
+
" border-color: transparent;\n",
|
396 |
+
" border-right-color: var(--fill-color);\n",
|
397 |
+
" border-bottom-color: var(--fill-color);\n",
|
398 |
+
" }\n",
|
399 |
+
" 90% {\n",
|
400 |
+
" border-color: transparent;\n",
|
401 |
+
" border-bottom-color: var(--fill-color);\n",
|
402 |
+
" }\n",
|
403 |
+
" }\n",
|
404 |
+
"</style>\n",
|
405 |
+
"\n",
|
406 |
+
" <script>\n",
|
407 |
+
" async function quickchart(key) {\n",
|
408 |
+
" const quickchartButtonEl =\n",
|
409 |
+
" document.querySelector('#' + key + ' button');\n",
|
410 |
+
" quickchartButtonEl.disabled = true; // To prevent multiple clicks.\n",
|
411 |
+
" quickchartButtonEl.classList.add('colab-df-spinner');\n",
|
412 |
+
" try {\n",
|
413 |
+
" const charts = await google.colab.kernel.invokeFunction(\n",
|
414 |
+
" 'suggestCharts', [key], {});\n",
|
415 |
+
" } catch (error) {\n",
|
416 |
+
" console.error('Error during call to suggestCharts:', error);\n",
|
417 |
+
" }\n",
|
418 |
+
" quickchartButtonEl.classList.remove('colab-df-spinner');\n",
|
419 |
+
" quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
|
420 |
+
" }\n",
|
421 |
+
" (() => {\n",
|
422 |
+
" let quickchartButtonEl =\n",
|
423 |
+
" document.querySelector('#df-cc66d435-72a8-4146-ba22-ea8ef5610445 button');\n",
|
424 |
+
" quickchartButtonEl.style.display =\n",
|
425 |
+
" google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
|
426 |
+
" })();\n",
|
427 |
+
" </script>\n",
|
428 |
+
"</div>\n",
|
429 |
+
"\n",
|
430 |
+
" </div>\n",
|
431 |
+
" </div>\n"
|
432 |
+
],
|
433 |
+
"text/plain": [
|
434 |
+
" deceptive polarity\n",
|
435 |
+
"count 1600.000000 1600.000000\n",
|
436 |
+
"mean 0.500000 0.500000\n",
|
437 |
+
"std 0.500156 0.500156\n",
|
438 |
+
"min 0.000000 0.000000\n",
|
439 |
+
"25% 0.000000 0.000000\n",
|
440 |
+
"50% 0.500000 0.500000\n",
|
441 |
+
"75% 1.000000 1.000000\n",
|
442 |
+
"max 1.000000 1.000000"
|
443 |
+
]
|
444 |
+
},
|
445 |
+
"execution_count": 15,
|
446 |
+
"metadata": {},
|
447 |
+
"output_type": "execute_result"
|
448 |
+
}
|
449 |
+
],
|
450 |
+
"source": [
|
451 |
+
"df = data.sample(frac=1)\n",
|
452 |
+
"df.describe()"
|
453 |
+
]
|
454 |
+
},
|
455 |
+
{
|
456 |
+
"cell_type": "code",
|
457 |
+
"execution_count": null,
|
458 |
+
"metadata": {
|
459 |
+
"id": "gqCqI9xSriAb"
|
460 |
+
},
|
461 |
+
"outputs": [],
|
462 |
+
"source": [
|
463 |
+
"def create_class(c):\n",
|
464 |
+
" if c['polarity'] == 1 and c['deceptive'] == 1:\n",
|
465 |
+
" return [1,1]\n",
|
466 |
+
" elif c['polarity'] == 1 and c['deceptive'] == 0:\n",
|
467 |
+
" return [1,0]\n",
|
468 |
+
" elif c['polarity'] == 0 and c['deceptive'] == 1:\n",
|
469 |
+
" return [0,1]\n",
|
470 |
+
" else:\n",
|
471 |
+
" return [0,0]\n",
|
472 |
+
"\n",
|
473 |
+
"def specific_class(c):\n",
|
474 |
+
" if c['polarity'] == 1 and c['deceptive'] == 1: # Actually Deceptive ---> 0\n",
|
475 |
+
" return \"TRUE_POSITIVE\"\n",
|
476 |
+
" elif c['polarity'] == 1 and c['deceptive'] == 0: # Actually Not Deceptive ---> 1\n",
|
477 |
+
" return \"FALSE_POSITIVE\"\n",
|
478 |
+
" elif c['polarity'] == 0 and c['deceptive'] == 1: # Actually Not Deceptive ---> 2\n",
|
479 |
+
" return \"TRUE_NEGATIVE\"\n",
|
480 |
+
" else: # Actually Deceptive ---> 3\n",
|
481 |
+
" return \"FALSE_NEGATIVE\"\n",
|
482 |
+
"\n",
|
483 |
+
"data['final_class'] = data.apply(create_class, axis=1)\n",
|
484 |
+
"data['given_class'] = data.apply(specific_class, axis=1)"
|
485 |
+
]
|
486 |
+
},
|
487 |
+
{
|
488 |
+
"cell_type": "code",
|
489 |
+
"execution_count": null,
|
490 |
+
"metadata": {
|
491 |
+
"id": "0KtN7332rkOJ"
|
492 |
+
},
|
493 |
+
"outputs": [],
|
494 |
+
"source": [
|
495 |
+
"from sklearn import preprocessing\n",
|
496 |
+
"\n",
|
497 |
+
"label_encoder = preprocessing.LabelEncoder()\n",
|
498 |
+
"\n",
|
499 |
+
"data['given_class'] = label_encoder.fit_transform(data['given_class'])"
|
500 |
+
]
|
501 |
+
},
|
502 |
+
{
|
503 |
+
"cell_type": "code",
|
504 |
+
"execution_count": 251,
|
505 |
+
"metadata": {
|
506 |
+
"colab": {
|
507 |
+
"base_uri": "https://localhost:8080/"
|
508 |
+
},
|
509 |
+
"id": "MW0O6_v3EM2G",
|
510 |
+
"outputId": "ab5238a9-8afb-4af5-8c49-69bab79c8caa"
|
511 |
+
},
|
512 |
+
"outputs": [
|
513 |
+
{
|
514 |
+
"data": {
|
515 |
+
"text/plain": [
|
516 |
+
"0 [1, 1]\n",
|
517 |
+
"1 [1, 1]\n",
|
518 |
+
"2 [1, 1]\n",
|
519 |
+
"3 [1, 1]\n",
|
520 |
+
"4 [1, 1]\n",
|
521 |
+
" ... \n",
|
522 |
+
"1595 [0, 0]\n",
|
523 |
+
"1596 [0, 0]\n",
|
524 |
+
"1597 [0, 0]\n",
|
525 |
+
"1598 [0, 0]\n",
|
526 |
+
"1599 [0, 0]\n",
|
527 |
+
"Name: final_class, Length: 1600, dtype: object"
|
528 |
+
]
|
529 |
+
},
|
530 |
+
"execution_count": 251,
|
531 |
+
"metadata": {},
|
532 |
+
"output_type": "execute_result"
|
533 |
+
}
|
534 |
+
],
|
535 |
+
"source": [
|
536 |
+
"data['final_class']"
|
537 |
+
]
|
538 |
+
},
|
539 |
+
{
|
540 |
+
"cell_type": "code",
|
541 |
+
"execution_count": null,
|
542 |
+
"metadata": {
|
543 |
+
"id": "F24YJHdermPy"
|
544 |
+
},
|
545 |
+
"outputs": [],
|
546 |
+
"source": [
|
547 |
+
"Y = data['given_class']\n",
|
548 |
+
"encoder = LabelEncoder()\n",
|
549 |
+
"encoder.fit(Y)\n",
|
550 |
+
"encoded_Y = encoder.transform(Y)\n",
|
551 |
+
"dummy_y = to_categorical(encoded_Y)"
|
552 |
+
]
|
553 |
+
},
|
554 |
+
{
|
555 |
+
"cell_type": "code",
|
556 |
+
"execution_count": 247,
|
557 |
+
"metadata": {
|
558 |
+
"colab": {
|
559 |
+
"base_uri": "https://localhost:8080/"
|
560 |
+
},
|
561 |
+
"id": "wmCJ8i83D1x4",
|
562 |
+
"outputId": "89d49003-3a75-421d-884d-e0d032cab6ca"
|
563 |
+
},
|
564 |
+
"outputs": [
|
565 |
+
{
|
566 |
+
"data": {
|
567 |
+
"text/plain": [
|
568 |
+
"array([[0., 0., 0., 1.],\n",
|
569 |
+
" [0., 0., 0., 1.],\n",
|
570 |
+
" [0., 0., 0., 1.],\n",
|
571 |
+
" ...,\n",
|
572 |
+
" [1., 0., 0., 0.],\n",
|
573 |
+
" [1., 0., 0., 0.],\n",
|
574 |
+
" [1., 0., 0., 0.]], dtype=float32)"
|
575 |
+
]
|
576 |
+
},
|
577 |
+
"execution_count": 247,
|
578 |
+
"metadata": {},
|
579 |
+
"output_type": "execute_result"
|
580 |
+
}
|
581 |
+
],
|
582 |
+
"source": [
|
583 |
+
"dummy_y"
|
584 |
+
]
|
585 |
+
},
|
586 |
+
{
|
587 |
+
"cell_type": "code",
|
588 |
+
"execution_count": null,
|
589 |
+
"metadata": {
|
590 |
+
"id": "U0OJ5Qhbrocj"
|
591 |
+
},
|
592 |
+
"outputs": [],
|
593 |
+
"source": [
|
594 |
+
"textData = pd.DataFrame(list(data['text']))\n"
|
595 |
+
]
|
596 |
+
},
|
597 |
+
{
|
598 |
+
"cell_type": "code",
|
599 |
+
"execution_count": null,
|
600 |
+
"metadata": {
|
601 |
+
"id": "c_WdJiovrwpN"
|
602 |
+
},
|
603 |
+
"outputs": [],
|
604 |
+
"source": [
|
605 |
+
"stemmer = snowballstemmer.EnglishStemmer()\n",
|
606 |
+
"\n",
|
607 |
+
"stop = stopwords.words('english')\n",
|
608 |
+
"stop.extend(['may','also','zero','one','two','three','four','five','six','seven','eight','nine','ten','across','among','beside','however','yet','within']+list(ascii_lowercase))\n",
|
609 |
+
"stoplist = stemmer.stemWords(stop)\n",
|
610 |
+
"stoplist = set(stoplist)\n",
|
611 |
+
"stop = set(sorted(stop + list(stoplist)))"
|
612 |
+
]
|
613 |
+
},
|
614 |
+
{
|
615 |
+
"cell_type": "code",
|
616 |
+
"execution_count": null,
|
617 |
+
"metadata": {
|
618 |
+
"colab": {
|
619 |
+
"base_uri": "https://localhost:8080/"
|
620 |
+
},
|
621 |
+
"id": "pUJtMqjZryE9",
|
622 |
+
"outputId": "295b19f7-94fa-447b-b964-4017b0593401"
|
623 |
+
},
|
624 |
+
"outputs": [
|
625 |
+
{
|
626 |
+
"name": "stderr",
|
627 |
+
"output_type": "stream",
|
628 |
+
"text": [
|
629 |
+
"[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
|
630 |
+
"[nltk_data] Unzipping corpora/stopwords.zip.\n"
|
631 |
+
]
|
632 |
+
},
|
633 |
+
{
|
634 |
+
"data": {
|
635 |
+
"text/plain": [
|
636 |
+
"True"
|
637 |
+
]
|
638 |
+
},
|
639 |
+
"execution_count": 23,
|
640 |
+
"metadata": {},
|
641 |
+
"output_type": "execute_result"
|
642 |
+
}
|
643 |
+
],
|
644 |
+
"source": [
|
645 |
+
"nltk.download('stopwords')"
|
646 |
+
]
|
647 |
+
},
|
648 |
+
{
|
649 |
+
"cell_type": "code",
|
650 |
+
"execution_count": null,
|
651 |
+
"metadata": {
|
652 |
+
"id": "_FG8kbvgr9HS"
|
653 |
+
},
|
654 |
+
"outputs": [],
|
655 |
+
"source": [
|
656 |
+
"textData[0].replace('[!\"#%\\'()*+,-./:;<=>?@\\[\\]^_`{|}~1234567890’”“′‘\\\\\\]',' ',inplace=True,regex=True)\n",
|
657 |
+
"wordlist = filter(None, \" \".join(list(set(list(itertools.chain(*textData[0].str.split(' ')))))).split(\" \"))\n",
|
658 |
+
"data['stemmed_text_data'] = [' '.join(filter(None,filter(lambda word: word not in stop, line))) for line in textData[0].str.lower().str.split(' ')]"
|
659 |
+
]
|
660 |
+
},
|
661 |
+
{
|
662 |
+
"cell_type": "code",
|
663 |
+
"execution_count": null,
|
664 |
+
"metadata": {
|
665 |
+
"id": "fjUfTsjEsDh2"
|
666 |
+
},
|
667 |
+
"outputs": [],
|
668 |
+
"source": [
|
669 |
+
"minimum_count = 1\n",
|
670 |
+
"str_frequencies = pd.DataFrame(list(Counter(filter(None,list(itertools.chain(*data['stemmed_text_data'].str.split(' '))))).items()),columns=['word','count'])\n",
|
671 |
+
"low_frequency_words = set(str_frequencies[str_frequencies['count'] < minimum_count]['word'])\n",
|
672 |
+
"data['stemmed_text_data'] = [' '.join(filter(None,filter(lambda word: word not in low_frequency_words, line))) for line in data['stemmed_text_data'].str.split(' ')]\n",
|
673 |
+
"data['stemmed_text_data'] = [\" \".join(stemmer.stemWords(re.sub('[!\"#%\\'()*+,-./:;<=>?@\\[\\]^_`{|}~1234567890’”“′‘\\\\\\]',' ', next_text).split(' '))) for next_text in data['stemmed_text_data']]"
|
674 |
+
]
|
675 |
+
},
|
676 |
+
{
|
677 |
+
"cell_type": "code",
|
678 |
+
"execution_count": null,
|
679 |
+
"metadata": {
|
680 |
+
"id": "GX-Gd8M6sEpp"
|
681 |
+
},
|
682 |
+
"outputs": [],
|
683 |
+
"source": [
|
684 |
+
"lmtzr = WordNetLemmatizer()\n",
|
685 |
+
"w = re.compile(\"\\w+\",re.I)\n",
|
686 |
+
"\n",
|
687 |
+
"def label_sentences(df, input_point):\n",
|
688 |
+
" labeled_sentences = []\n",
|
689 |
+
" list_sen = []\n",
|
690 |
+
" for index, datapoint in df.iterrows():\n",
|
691 |
+
" tokenized_words = re.findall(w,datapoint[input_point].lower())\n",
|
692 |
+
" labeled_sentences.append(TaggedDocument(words=tokenized_words, tags=['SENT_%s' %index]))\n",
|
693 |
+
" list_sen.append(tokenized_words)\n",
|
694 |
+
" return labeled_sentences, list_sen\n",
|
695 |
+
"\n",
|
696 |
+
"def train_doc2vec_model(labeled_sentences):\n",
|
697 |
+
" model = Doc2Vec(min_count=1, window=9, vector_size=512, sample=1e-4, negative=5, workers=7)\n",
|
698 |
+
" model.build_vocab(labeled_sentences)\n",
|
699 |
+
" pretrained_weights = model.wv.vectors\n",
|
700 |
+
" vocab_size, embedding_size = pretrained_weights.shape\n",
|
701 |
+
" model.train(labeled_sentences, total_examples=vocab_size, epochs=400)\n",
|
702 |
+
"\n",
|
703 |
+
" return model"
|
704 |
+
]
|
705 |
+
},
|
706 |
+
{
|
707 |
+
"cell_type": "code",
|
708 |
+
"execution_count": null,
|
709 |
+
"metadata": {
|
710 |
+
"id": "2C_5s3UOsGfU"
|
711 |
+
},
|
712 |
+
"outputs": [],
|
713 |
+
"source": [
|
714 |
+
"textData = data['stemmed_text_data'].to_frame().reset_index()\n",
|
715 |
+
"sen, corpus = label_sentences(textData, 'stemmed_text_data')"
|
716 |
+
]
|
717 |
+
},
|
718 |
+
{
|
719 |
+
"cell_type": "code",
|
720 |
+
"execution_count": null,
|
721 |
+
"metadata": {
|
722 |
+
"id": "qpb3aMvW_jdj"
|
723 |
+
},
|
724 |
+
"outputs": [],
|
725 |
+
"source": [
|
726 |
+
"sen"
|
727 |
+
]
|
728 |
+
},
|
729 |
+
{
|
730 |
+
"cell_type": "code",
|
731 |
+
"execution_count": null,
|
732 |
+
"metadata": {
|
733 |
+
"id": "JeK3t6_HsNv9"
|
734 |
+
},
|
735 |
+
"outputs": [],
|
736 |
+
"source": [
|
737 |
+
"doc2vec_model = train_doc2vec_model(sen)"
|
738 |
+
]
|
739 |
+
},
|
740 |
+
{
|
741 |
+
"cell_type": "code",
|
742 |
+
"execution_count": null,
|
743 |
+
"metadata": {
|
744 |
+
"id": "DyX1XG1usQMm"
|
745 |
+
},
|
746 |
+
"outputs": [],
|
747 |
+
"source": [
|
748 |
+
"doc2vec_model.save(\"doc2vec_model_opinion_corpus.d2v\")"
|
749 |
+
]
|
750 |
+
},
|
751 |
+
{
|
752 |
+
"cell_type": "code",
|
753 |
+
"execution_count": null,
|
754 |
+
"metadata": {
|
755 |
+
"id": "l0OAFencszum"
|
756 |
+
},
|
757 |
+
"outputs": [],
|
758 |
+
"source": [
|
759 |
+
"doc2vec_model = Doc2Vec.load(\"doc2vec_model_opinion_corpus.d2v\")"
|
760 |
+
]
|
761 |
+
},
|
762 |
+
{
|
763 |
+
"cell_type": "code",
|
764 |
+
"execution_count": null,
|
765 |
+
"metadata": {
|
766 |
+
"colab": {
|
767 |
+
"base_uri": "https://localhost:8080/"
|
768 |
+
},
|
769 |
+
"id": "ZJTp1POIs1sQ",
|
770 |
+
"outputId": "5788b8b8-3f66-4d4a-e22a-6bff020adf1e"
|
771 |
+
},
|
772 |
+
"outputs": [
|
773 |
+
{
|
774 |
+
"name": "stderr",
|
775 |
+
"output_type": "stream",
|
776 |
+
"text": [
|
777 |
+
"/usr/local/lib/python3.10/dist-packages/sklearn/feature_extraction/text.py:528: UserWarning: The parameter 'token_pattern' will not be used since 'tokenizer' is not None'\n",
|
778 |
+
" warnings.warn(\n"
|
779 |
+
]
|
780 |
+
}
|
781 |
+
],
|
782 |
+
"source": [
|
783 |
+
"from sklearn.feature_extraction.text import TfidfVectorizer\n",
|
784 |
+
"from sklearn.decomposition import TruncatedSVD\n",
|
785 |
+
"\n",
|
786 |
+
"tfidf1 = TfidfVectorizer(tokenizer=lambda i:i, lowercase=False, ngram_range=(1,1))\n",
|
787 |
+
"result_train1 = tfidf1.fit_transform(corpus)\n",
|
788 |
+
"\n",
|
789 |
+
"tfidf2 = TfidfVectorizer(tokenizer=lambda i:i, lowercase=False, ngram_range=(1,2))\n",
|
790 |
+
"result_train2 = tfidf2.fit_transform(corpus)\n",
|
791 |
+
"\n",
|
792 |
+
"tfidf3 = TfidfVectorizer(tokenizer=lambda i:i, lowercase=False, ngram_range=(1,3))\n",
|
793 |
+
"result_train3 = tfidf3.fit_transform(corpus)\n",
|
794 |
+
"\n",
|
795 |
+
"svd = TruncatedSVD(n_components=512, n_iter=40, random_state=34)\n",
|
796 |
+
"tfidf_data1 = svd.fit_transform(result_train1)\n",
|
797 |
+
"tfidf_data2 = svd.fit_transform(result_train2)\n",
|
798 |
+
"tfidf_data3 = svd.fit_transform(result_train3)"
|
799 |
+
]
|
800 |
+
},
|
801 |
+
{
|
802 |
+
"cell_type": "code",
|
803 |
+
"execution_count": null,
|
804 |
+
"metadata": {
|
805 |
+
"id": "io0D71F00Wv8"
|
806 |
+
},
|
807 |
+
"outputs": [],
|
808 |
+
"source": [
|
809 |
+
"nlp = spacy.load(\"en_core_web_sm\")"
|
810 |
+
]
|
811 |
+
},
|
812 |
+
{
|
813 |
+
"cell_type": "code",
|
814 |
+
"execution_count": null,
|
815 |
+
"metadata": {
|
816 |
+
"id": "QR6PqZREs3EA"
|
817 |
+
},
|
818 |
+
"outputs": [],
|
819 |
+
"source": [
|
820 |
+
"from sklearn.feature_extraction.text import CountVectorizer\n",
|
821 |
+
"import spacy\n",
|
822 |
+
"\n",
|
823 |
+
"nlp = spacy.load(\"en_core_web_sm\")\n",
|
824 |
+
"temp_textData = pd.DataFrame(list(data['text']))\n",
|
825 |
+
"\n",
|
826 |
+
"overall_pos_tags_tokens = []\n",
|
827 |
+
"overall_pos = []\n",
|
828 |
+
"overall_tokens = []\n",
|
829 |
+
"overall_dep = []\n",
|
830 |
+
"\n",
|
831 |
+
"for i in range(1600):\n",
|
832 |
+
" doc = nlp(temp_textData[0][i])\n",
|
833 |
+
" given_pos_tags_tokens = []\n",
|
834 |
+
" given_pos = []\n",
|
835 |
+
" given_tokens = []\n",
|
836 |
+
" given_dep = []\n",
|
837 |
+
" for token in doc:\n",
|
838 |
+
" output = \"%s_%s\" % (token.pos_, token.tag_)\n",
|
839 |
+
" given_pos_tags_tokens.append(output)\n",
|
840 |
+
" given_pos.append(token.pos_)\n",
|
841 |
+
" given_tokens.append(token.tag_)\n",
|
842 |
+
" given_dep.append(token.dep_)\n",
|
843 |
+
"\n",
|
844 |
+
" overall_pos_tags_tokens.append(given_pos_tags_tokens)\n",
|
845 |
+
" overall_pos.append(given_pos)\n",
|
846 |
+
" overall_tokens.append(given_tokens)\n",
|
847 |
+
" overall_dep.append(given_dep)\n"
|
848 |
+
]
|
849 |
+
},
|
850 |
+
{
|
851 |
+
"cell_type": "code",
|
852 |
+
"execution_count": null,
|
853 |
+
"metadata": {
|
854 |
+
"id": "O6C2OJ8KEzHk"
|
855 |
+
},
|
856 |
+
"outputs": [],
|
857 |
+
"source": [
|
858 |
+
"overall_tokens"
|
859 |
+
]
|
860 |
+
},
|
861 |
+
{
|
862 |
+
"cell_type": "code",
|
863 |
+
"execution_count": null,
|
864 |
+
"metadata": {
|
865 |
+
"colab": {
|
866 |
+
"base_uri": "https://localhost:8080/"
|
867 |
+
},
|
868 |
+
"id": "4PxkBoQgs4wV",
|
869 |
+
"outputId": "07f59ab7-dbc0-4b41-e712-31a165e6ddf2"
|
870 |
+
},
|
871 |
+
"outputs": [
|
872 |
+
{
|
873 |
+
"name": "stderr",
|
874 |
+
"output_type": "stream",
|
875 |
+
"text": [
|
876 |
+
"/usr/local/lib/python3.10/dist-packages/sklearn/feature_extraction/text.py:528: UserWarning: The parameter 'token_pattern' will not be used since 'tokenizer' is not None'\n",
|
877 |
+
" warnings.warn(\n"
|
878 |
+
]
|
879 |
+
}
|
880 |
+
],
|
881 |
+
"source": [
|
882 |
+
"import numpy as np\n",
|
883 |
+
"from sklearn.feature_extraction.text import CountVectorizer\n",
|
884 |
+
"from sklearn.preprocessing import MinMaxScaler\n",
|
885 |
+
"\n",
|
886 |
+
"count = CountVectorizer(tokenizer=lambda i: i, lowercase=False)\n",
|
887 |
+
"pos_tags_data = count.fit_transform(overall_pos_tags_tokens).todense()\n",
|
888 |
+
"pos_data = count.fit_transform(overall_pos).todense()\n",
|
889 |
+
"tokens_data = count.fit_transform(overall_tokens).todense()\n",
|
890 |
+
"dep_data = count.fit_transform(overall_dep).todense()\n",
|
891 |
+
"\n",
|
892 |
+
"min_max_scaler = MinMaxScaler()\n",
|
893 |
+
"\n",
|
894 |
+
"normalized_pos_tags_data = min_max_scaler.fit_transform(np.asarray(pos_tags_data))\n",
|
895 |
+
"normalized_pos_data = min_max_scaler.fit_transform(np.asarray(pos_data))\n",
|
896 |
+
"normalized_tokens_data = min_max_scaler.fit_transform(np.asarray(tokens_data))\n",
|
897 |
+
"normalized_dep_data = min_max_scaler.fit_transform(np.asarray(dep_data))\n",
|
898 |
+
"\n",
|
899 |
+
"# Convert the scaled data to numpy arrays\n",
|
900 |
+
"normalized_pos_tags_data = np.asarray(normalized_pos_tags_data)\n",
|
901 |
+
"normalized_pos_data = np.asarray(normalized_pos_data)\n",
|
902 |
+
"normalized_tokens_data = np.asarray(normalized_tokens_data)\n",
|
903 |
+
"normalized_dep_data = np.asarray(normalized_dep_data)\n",
|
904 |
+
"\n",
|
905 |
+
"final_pos_tags_data = np.zeros(shape=(1600, 512)).astype(np.float32)\n",
|
906 |
+
"final_pos_data = np.zeros(shape=(1600, 512)).astype(np.float32)\n",
|
907 |
+
"final_tokens_data = np.zeros(shape=(1600, 512)).astype(np.float32)\n",
|
908 |
+
"final_dep_data = np.zeros(shape=(1600, 512)).astype(np.float32)\n",
|
909 |
+
"\n",
|
910 |
+
"# Assign the converted arrays to the final arrays\n",
|
911 |
+
"final_pos_tags_data[:normalized_pos_tags_data.shape[0], :normalized_pos_tags_data.shape[1]] = normalized_pos_tags_data\n",
|
912 |
+
"final_pos_data[:normalized_pos_data.shape[0], :normalized_pos_data.shape[1]] = normalized_pos_data\n",
|
913 |
+
"final_tokens_data[:normalized_tokens_data.shape[0], :normalized_tokens_data.shape[1]] = normalized_tokens_data\n",
|
914 |
+
"final_dep_data[:normalized_dep_data.shape[0], :normalized_dep_data.shape[1]] = normalized_dep_data\n"
|
915 |
+
]
|
916 |
+
},
|
917 |
+
{
|
918 |
+
"cell_type": "code",
|
919 |
+
"execution_count": null,
|
920 |
+
"metadata": {
|
921 |
+
"colab": {
|
922 |
+
"base_uri": "https://localhost:8080/"
|
923 |
+
},
|
924 |
+
"id": "jQdeLCgas6HD",
|
925 |
+
"outputId": "b1079e7c-c4fa-413d-bf89-9f5dc8fe36d8"
|
926 |
+
},
|
927 |
+
"outputs": [
|
928 |
+
{
|
929 |
+
"name": "stdout",
|
930 |
+
"output_type": "stream",
|
931 |
+
"text": [
|
932 |
+
"370\n"
|
933 |
+
]
|
934 |
+
}
|
935 |
+
],
|
936 |
+
"source": [
|
937 |
+
"maxlength = []\n",
|
938 |
+
"for i in range(0,len(sen)):\n",
|
939 |
+
" maxlength.append(len(sen[i][0]))\n",
|
940 |
+
"\n",
|
941 |
+
"print(max(maxlength))"
|
942 |
+
]
|
943 |
+
},
|
944 |
+
{
|
945 |
+
"cell_type": "code",
|
946 |
+
"execution_count": null,
|
947 |
+
"metadata": {
|
948 |
+
"id": "X5y1kjW-s7bJ"
|
949 |
+
},
|
950 |
+
"outputs": [],
|
951 |
+
"source": [
|
952 |
+
"doc2vec_model = Doc2Vec.load(\"doc2vec_model_opinion_corpus.d2v\")\n",
|
953 |
+
"\n",
|
954 |
+
"def vectorize_comments(df,d2v_model):\n",
|
955 |
+
" y = []\n",
|
956 |
+
" comments = []\n",
|
957 |
+
" for i in range(0,df.shape[0]):\n",
|
958 |
+
" label = 'SENT_%s' %i\n",
|
959 |
+
" comments.append(d2v_model.docvecs[label])\n",
|
960 |
+
" df['vectorized_comments'] = comments\n",
|
961 |
+
"\n",
|
962 |
+
" return df\n",
|
963 |
+
"\n",
|
964 |
+
"textData = vectorize_comments(textData,doc2vec_model)\n",
|
965 |
+
"print (textData.head(2))"
|
966 |
+
]
|
967 |
+
},
|
968 |
+
{
|
969 |
+
"cell_type": "code",
|
970 |
+
"execution_count": null,
|
971 |
+
"metadata": {
|
972 |
+
"id": "SUfiSDENs8cg"
|
973 |
+
},
|
974 |
+
"outputs": [],
|
975 |
+
"source": [
|
976 |
+
"from sklearn.model_selection import train_test_split\n",
|
977 |
+
"from sklearn.model_selection import cross_validate,GridSearchCV\n",
|
978 |
+
"\n",
|
979 |
+
"X_train, X_test, y_train, y_test = train_test_split(textData[\"vectorized_comments\"].T.tolist(),\n",
|
980 |
+
" dummy_y,\n",
|
981 |
+
" test_size=0.1,\n",
|
982 |
+
" random_state=56)"
|
983 |
+
]
|
984 |
+
},
|
985 |
+
{
|
986 |
+
"cell_type": "code",
|
987 |
+
"execution_count": null,
|
988 |
+
"metadata": {
|
989 |
+
"id": "1ID0T5d0s-iS"
|
990 |
+
},
|
991 |
+
"outputs": [],
|
992 |
+
"source": [
|
993 |
+
"X = np.array(textData[\"vectorized_comments\"].T.tolist()).reshape((1,1600,512))\n",
|
994 |
+
"y = np.array(dummy_y).reshape((1600,4))\n",
|
995 |
+
"X_train2 = np.array(X_train).reshape((1,1440,512))\n",
|
996 |
+
"y_train2 = np.array(y_train).reshape((1,1440,4))\n",
|
997 |
+
"X_test2 = np.array(X_test).reshape((1,160,512))\n",
|
998 |
+
"y_test2 = np.array(y_test).reshape((1,160,4))"
|
999 |
+
]
|
1000 |
+
},
|
1001 |
+
{
|
1002 |
+
"cell_type": "code",
|
1003 |
+
"execution_count": null,
|
1004 |
+
"metadata": {
|
1005 |
+
"id": "GlYOfPhGs_lR"
|
1006 |
+
},
|
1007 |
+
"outputs": [],
|
1008 |
+
"source": [
|
1009 |
+
"from sklearn.model_selection import StratifiedKFold\n",
|
1010 |
+
"Xtemp = df[\"vectorized_comments\"].T.tolist()\n",
|
1011 |
+
"ytemp = data['given_class']\n",
|
1012 |
+
"training_indices = []\n",
|
1013 |
+
"testing_indices = []\n",
|
1014 |
+
"\n",
|
1015 |
+
"skf = StratifiedKFold(n_splits=10)\n",
|
1016 |
+
"skf.get_n_splits(Xtemp, ytemp)\n",
|
1017 |
+
"\n",
|
1018 |
+
"for train_index, test_index in skf.split(Xtemp, ytemp):\n",
|
1019 |
+
" training_indices.append(train_index)\n",
|
1020 |
+
" testing_indices.append(test_index)"
|
1021 |
+
]
|
1022 |
+
},
|
1023 |
+
{
|
1024 |
+
"cell_type": "code",
|
1025 |
+
"execution_count": 238,
|
1026 |
+
"metadata": {
|
1027 |
+
"colab": {
|
1028 |
+
"base_uri": "https://localhost:8080/"
|
1029 |
+
},
|
1030 |
+
"id": "-avX-WdT_Z2P",
|
1031 |
+
"outputId": "a2a227a5-9ff7-4ee0-8cd0-d6b8fb157363"
|
1032 |
+
},
|
1033 |
+
"outputs": [
|
1034 |
+
{
|
1035 |
+
"data": {
|
1036 |
+
"text/plain": [
|
1037 |
+
"160"
|
1038 |
+
]
|
1039 |
+
},
|
1040 |
+
"execution_count": 238,
|
1041 |
+
"metadata": {},
|
1042 |
+
"output_type": "execute_result"
|
1043 |
+
}
|
1044 |
+
],
|
1045 |
+
"source": [
|
1046 |
+
"len(testing_indices[2])"
|
1047 |
+
]
|
1048 |
+
},
|
1049 |
+
{
|
1050 |
+
"cell_type": "code",
|
1051 |
+
"execution_count": null,
|
1052 |
+
"metadata": {
|
1053 |
+
"id": "EIVEsyLiRjn7"
|
1054 |
+
},
|
1055 |
+
"outputs": [],
|
1056 |
+
"source": [
|
1057 |
+
"training_indices"
|
1058 |
+
]
|
1059 |
+
},
|
1060 |
+
{
|
1061 |
+
"cell_type": "code",
|
1062 |
+
"execution_count": 211,
|
1063 |
+
"metadata": {
|
1064 |
+
"colab": {
|
1065 |
+
"base_uri": "https://localhost:8080/"
|
1066 |
+
},
|
1067 |
+
"id": "iPc15Lwv7-L2",
|
1068 |
+
"outputId": "5b5578e0-27b7-4bad-92e7-ef42dda066ed"
|
1069 |
+
},
|
1070 |
+
"outputs": [
|
1071 |
+
{
|
1072 |
+
"data": {
|
1073 |
+
"text/plain": [
|
1074 |
+
"<12x12 sparse matrix of type '<class 'numpy.float64'>'\n",
|
1075 |
+
"\twith 20 stored elements in Compressed Sparse Row format>"
|
1076 |
+
]
|
1077 |
+
},
|
1078 |
+
"execution_count": 211,
|
1079 |
+
"metadata": {},
|
1080 |
+
"output_type": "execute_result"
|
1081 |
+
}
|
1082 |
+
],
|
1083 |
+
"source": [
|
1084 |
+
"result_train1"
|
1085 |
+
]
|
1086 |
+
},
|
1087 |
+
{
|
1088 |
+
"cell_type": "code",
|
1089 |
+
"execution_count": 212,
|
1090 |
+
"metadata": {
|
1091 |
+
"id": "8UwiqmnhtAz6"
|
1092 |
+
},
|
1093 |
+
"outputs": [],
|
1094 |
+
"source": [
|
1095 |
+
"def extractTrainingAndTestingData(givenIndex):\n",
|
1096 |
+
" X_train3 = np.zeros(shape=(1440, max(maxlength)+10, 512)).astype(np.float32)\n",
|
1097 |
+
" Y_train3 = np.zeros(shape=(1440, 4)).astype(np.float32)\n",
|
1098 |
+
" X_test3 = np.zeros(shape=(160, max(maxlength)+10, 512)).astype(np.float32)\n",
|
1099 |
+
" Y_test3 = np.zeros(shape=(160, 4)).astype(np.float32)\n",
|
1100 |
+
"\n",
|
1101 |
+
" empty_word = np.zeros(512).astype(np.float32)\n",
|
1102 |
+
"\n",
|
1103 |
+
" count_i = 0\n",
|
1104 |
+
" for i in training_indices[givenIndex]:\n",
|
1105 |
+
" len1 = len(sen[i][0])\n",
|
1106 |
+
" average_vector1 = np.zeros(512).astype(np.float32)\n",
|
1107 |
+
" average_vector2 = np.zeros(512).astype(np.float32)\n",
|
1108 |
+
" average_vector3 = np.zeros(512).astype(np.float32)\n",
|
1109 |
+
" for j in range(max(maxlength)+10):\n",
|
1110 |
+
" if j < len1:\n",
|
1111 |
+
" X_train3[count_i,j,:] = doc2vec_model[sen[i][0][j]]\n",
|
1112 |
+
" average_vector1 += result_train1[i, tfidf1.vocabulary_[sen[i][0][j]]] * doc2vec_model[sen[i][0][j]]\n",
|
1113 |
+
" average_vector2 += result_train2[i, tfidf2.vocabulary_[sen[i][0][j]]] * doc2vec_model[sen[i][0][j]]\n",
|
1114 |
+
" average_vector3 += result_train3[i, tfidf3.vocabulary_[sen[i][0][j]]] * doc2vec_model[sen[i][0][j]]\n",
|
1115 |
+
" #elif j >= len1 and j < len1 + 379:\n",
|
1116 |
+
" # X_train3[count_i,j,:] = glove_data[i, j-len1, :]\n",
|
1117 |
+
" elif j == len1:\n",
|
1118 |
+
" X_train3[count_i,j,:] = tfidf_data1[i]\n",
|
1119 |
+
" elif j == len1 + 1:\n",
|
1120 |
+
" X_train3[count_i,j,:] = tfidf_data2[i]\n",
|
1121 |
+
" elif j == len1+2:\n",
|
1122 |
+
" X_train3[count_i,j,:] = tfidf_data3[i]\n",
|
1123 |
+
" elif j == len1+3:\n",
|
1124 |
+
" X_train3[count_i,j,:] = average_vector1\n",
|
1125 |
+
" elif j == len1+4:\n",
|
1126 |
+
" X_train3[count_i,j,:] = average_vector2\n",
|
1127 |
+
" elif j == len1+5:\n",
|
1128 |
+
" X_train3[count_i,j,:] = average_vector3\n",
|
1129 |
+
" elif j == len1+6:\n",
|
1130 |
+
" X_train3[count_i,j,:] = final_pos_tags_data[i]\n",
|
1131 |
+
" elif j == len1+7:\n",
|
1132 |
+
" X_train3[count_i,j,:] = final_pos_data[i]\n",
|
1133 |
+
" elif j == len1+8:\n",
|
1134 |
+
" X_train3[count_i,j,:] = final_tokens_data[i]\n",
|
1135 |
+
" elif j == len1+9:\n",
|
1136 |
+
" X_train3[count_i,j,:] = final_dep_data[i]\n",
|
1137 |
+
" else:\n",
|
1138 |
+
" X_train3[count_i,j,:] = empty_word\n",
|
1139 |
+
"\n",
|
1140 |
+
" Y_train3[count_i,:] = dummy_y[i]\n",
|
1141 |
+
" count_i += 1\n",
|
1142 |
+
"\n",
|
1143 |
+
"\n",
|
1144 |
+
" count_i = 0\n",
|
1145 |
+
" for i in testing_indices[givenIndex]:\n",
|
1146 |
+
" len1 = len(sen[i][0])\n",
|
1147 |
+
" average_vector1 = np.zeros(512).astype(np.float32)\n",
|
1148 |
+
" average_vector2 = np.zeros(512).astype(np.float32)\n",
|
1149 |
+
" average_vector3 = np.zeros(512).astype(np.float32)\n",
|
1150 |
+
" for j in range(max(maxlength)+10):\n",
|
1151 |
+
" if j < len1:\n",
|
1152 |
+
" X_test3[count_i,j,:] = doc2vec_model[sen[i][0][j]]\n",
|
1153 |
+
" average_vector1 += result_train1[i, tfidf1.vocabulary_[sen[i][0][j]]] * doc2vec_model[sen[i][0][j]]\n",
|
1154 |
+
" average_vector2 += result_train2[i, tfidf2.vocabulary_[sen[i][0][j]]] * doc2vec_model[sen[i][0][j]]\n",
|
1155 |
+
" average_vector3 += result_train3[i, tfidf3.vocabulary_[sen[i][0][j]]] * doc2vec_model[sen[i][0][j]]\n",
|
1156 |
+
" #elif j >= len1 and j < len1 + 379:\n",
|
1157 |
+
" # X_test3[count_i,j,:] = glove_data[i, j-len1, :]\n",
|
1158 |
+
" elif j == len1:\n",
|
1159 |
+
" X_test3[count_i,j,:] = tfidf_data1[i]\n",
|
1160 |
+
" elif j == len1 + 1:\n",
|
1161 |
+
" X_test3[count_i,j,:] = tfidf_data2[i]\n",
|
1162 |
+
" elif j == len1+2:\n",
|
1163 |
+
" X_test3[count_i,j,:] = tfidf_data3[i]\n",
|
1164 |
+
" elif j == len1+3:\n",
|
1165 |
+
" X_test3[count_i,j,:] = average_vector1\n",
|
1166 |
+
" elif j == len1+4:\n",
|
1167 |
+
" X_test3[count_i,j,:] = average_vector2\n",
|
1168 |
+
" elif j == len1+5:\n",
|
1169 |
+
" X_test3[count_i,j,:] = average_vector3\n",
|
1170 |
+
" elif j == len1+6:\n",
|
1171 |
+
" X_test3[count_i,j,:] = final_pos_tags_data[i]\n",
|
1172 |
+
" elif j == len1+7:\n",
|
1173 |
+
" X_test3[count_i,j,:] = final_pos_data[i]\n",
|
1174 |
+
" elif j == len1+8:\n",
|
1175 |
+
" X_test3[count_i,j,:] = final_tokens_data[i]\n",
|
1176 |
+
" elif j == len1+9:\n",
|
1177 |
+
" X_test3[count_i,j,:] = final_dep_data[i]\n",
|
1178 |
+
" else:\n",
|
1179 |
+
" X_test3[count_i,j,:] = empty_word\n",
|
1180 |
+
"\n",
|
1181 |
+
" Y_test3[count_i,:] = dummy_y[i]\n",
|
1182 |
+
" count_i += 1\n",
|
1183 |
+
"\n",
|
1184 |
+
" return X_train3, X_test3, Y_train3, Y_test3"
|
1185 |
+
]
|
1186 |
+
},
|
1187 |
+
{
|
1188 |
+
"cell_type": "code",
|
1189 |
+
"execution_count": null,
|
1190 |
+
"metadata": {
|
1191 |
+
"id": "_ZQ6S5IhtB_8"
|
1192 |
+
},
|
1193 |
+
"outputs": [],
|
1194 |
+
"source": [
|
1195 |
+
"model = Sequential()\n",
|
1196 |
+
"model.add(Conv1D(filters=128, kernel_size=9, padding='same', activation='relu', input_shape=(max(maxlength)+10,512)))\n",
|
1197 |
+
"model.add(Dropout(0.25))\n",
|
1198 |
+
"model.add(MaxPooling1D(pool_size=2))\n",
|
1199 |
+
"model.add(Dropout(0.25))\n",
|
1200 |
+
"model.add(Conv1D(filters=128, kernel_size=7, padding='same', activation='relu'))\n",
|
1201 |
+
"model.add(Dropout(0.25))\n",
|
1202 |
+
"model.add(MaxPooling1D(pool_size=2))\n",
|
1203 |
+
"model.add(Dropout(0.25))\n",
|
1204 |
+
"model.add(Conv1D(filters=128, kernel_size=5, padding='same', activation='relu'))\n",
|
1205 |
+
"model.add(Dropout(0.25))\n",
|
1206 |
+
"model.add(Bidirectional(LSTM(50, dropout=0.25, recurrent_dropout=0.2)))\n",
|
1207 |
+
"model.add(Dense(4, activation='softmax'))\n",
|
1208 |
+
"model.compile(loss='binary_crossentropy', optimizer='Adam', metrics=['accuracy'])\n",
|
1209 |
+
"print(model.summary())"
|
1210 |
+
]
|
1211 |
+
},
|
1212 |
+
{
|
1213 |
+
"cell_type": "code",
|
1214 |
+
"execution_count": null,
|
1215 |
+
"metadata": {
|
1216 |
+
"id": "G2XuZvBOtDMs"
|
1217 |
+
},
|
1218 |
+
"outputs": [],
|
1219 |
+
"source": [
|
1220 |
+
"from sklearn.metrics import accuracy_score\n",
|
1221 |
+
"from keras.callbacks import ModelCheckpoint\n",
|
1222 |
+
"\n",
|
1223 |
+
"final_accuracies = []\n",
|
1224 |
+
"\n",
|
1225 |
+
"filename = 'weights.best.from_scratch%s.hdf5' % 9\n",
|
1226 |
+
"checkpointer = ModelCheckpoint(filepath=filename, verbose=1, save_best_only=True)\n",
|
1227 |
+
"X_train3, X_test3, Y_train3, Y_test3 = extractTrainingAndTestingData(9)"
|
1228 |
+
]
|
1229 |
+
},
|
1230 |
+
{
|
1231 |
+
"cell_type": "code",
|
1232 |
+
"execution_count": null,
|
1233 |
+
"metadata": {
|
1234 |
+
"id": "kGGf09dktEmS"
|
1235 |
+
},
|
1236 |
+
"outputs": [],
|
1237 |
+
"source": [
|
1238 |
+
"history = model.fit(X_train3, Y_train3, epochs=15, batch_size=512, callbacks=[checkpointer], validation_data=(X_test3, Y_test3), verbose=1)"
|
1239 |
+
]
|
1240 |
+
},
|
1241 |
+
{
|
1242 |
+
"cell_type": "code",
|
1243 |
+
"execution_count": null,
|
1244 |
+
"metadata": {
|
1245 |
+
"id": "ug5x2h7xtGNb"
|
1246 |
+
},
|
1247 |
+
"outputs": [],
|
1248 |
+
"source": [
|
1249 |
+
"model.evaluate(X_test3, Y_test3)"
|
1250 |
+
]
|
1251 |
+
},
|
1252 |
+
{
|
1253 |
+
"cell_type": "code",
|
1254 |
+
"execution_count": 207,
|
1255 |
+
"metadata": {
|
1256 |
+
"id": "UVKUsrk0tHil"
|
1257 |
+
},
|
1258 |
+
"outputs": [],
|
1259 |
+
"source": [
|
1260 |
+
"import matplotlib.pyplot as plt"
|
1261 |
+
]
|
1262 |
+
},
|
1263 |
+
{
|
1264 |
+
"cell_type": "code",
|
1265 |
+
"execution_count": null,
|
1266 |
+
"metadata": {
|
1267 |
+
"id": "BPcCy47XtKcL"
|
1268 |
+
},
|
1269 |
+
"outputs": [],
|
1270 |
+
"source": [
|
1271 |
+
"model.load_weights(filename)"
|
1272 |
+
]
|
1273 |
+
},
|
1274 |
+
{
|
1275 |
+
"cell_type": "code",
|
1276 |
+
"execution_count": null,
|
1277 |
+
"metadata": {
|
1278 |
+
"id": "csp-z21UtLUT"
|
1279 |
+
},
|
1280 |
+
"outputs": [],
|
1281 |
+
"source": [
|
1282 |
+
"for i in range(10):\n",
|
1283 |
+
" filename = 'weights.best.from_scratch%s.hdf5' % i\n",
|
1284 |
+
" checkpointer = ModelCheckpoint(filepath=filename, verbose=1, save_best_only=True)\n",
|
1285 |
+
" X_train3, X_test3, Y_train3, Y_test3 = extractTrainingAndTestingData(i)\n",
|
1286 |
+
" model.fit(X_train3, Y_train3, epochs=10, batch_size=512, callbacks=[checkpointer], validation_data=(X_test3, Y_test3))\n",
|
1287 |
+
" model.load_weights(filename)\n",
|
1288 |
+
" predicted = np.rint(model.predict(X_test3))\n",
|
1289 |
+
" final_accuracies.append(accuracy_score(Y_test3, predicted))\n",
|
1290 |
+
" print(accuracy_score(Y_test3, predicted))"
|
1291 |
+
]
|
1292 |
+
},
|
1293 |
+
{
|
1294 |
+
"cell_type": "code",
|
1295 |
+
"execution_count": null,
|
1296 |
+
"metadata": {
|
1297 |
+
"colab": {
|
1298 |
+
"base_uri": "https://localhost:8080/"
|
1299 |
+
},
|
1300 |
+
"id": "aKhWVJjFLE_9",
|
1301 |
+
"outputId": "0a8053a8-3bae-46b1-b154-31fd2030465b"
|
1302 |
+
},
|
1303 |
+
"outputs": [
|
1304 |
+
{
|
1305 |
+
"data": {
|
1306 |
+
"text/plain": [
|
1307 |
+
"380"
|
1308 |
+
]
|
1309 |
+
},
|
1310 |
+
"execution_count": 162,
|
1311 |
+
"metadata": {},
|
1312 |
+
"output_type": "execute_result"
|
1313 |
+
}
|
1314 |
+
],
|
1315 |
+
"source": [
|
1316 |
+
"len(X_test3[0])"
|
1317 |
+
]
|
1318 |
+
},
|
1319 |
+
{
|
1320 |
+
"cell_type": "code",
|
1321 |
+
"execution_count": null,
|
1322 |
+
"metadata": {
|
1323 |
+
"colab": {
|
1324 |
+
"base_uri": "https://localhost:8080/"
|
1325 |
+
},
|
1326 |
+
"id": "73d3h_lhL5r8",
|
1327 |
+
"outputId": "8da114d8-2901-49b2-fbcd-90b821b392ad"
|
1328 |
+
},
|
1329 |
+
"outputs": [
|
1330 |
+
{
|
1331 |
+
"data": {
|
1332 |
+
"text/plain": [
|
1333 |
+
"160"
|
1334 |
+
]
|
1335 |
+
},
|
1336 |
+
"execution_count": 161,
|
1337 |
+
"metadata": {},
|
1338 |
+
"output_type": "execute_result"
|
1339 |
+
}
|
1340 |
+
],
|
1341 |
+
"source": [
|
1342 |
+
"len(Y_test3)"
|
1343 |
+
]
|
1344 |
+
},
|
1345 |
+
{
|
1346 |
+
"cell_type": "code",
|
1347 |
+
"execution_count": null,
|
1348 |
+
"metadata": {
|
1349 |
+
"colab": {
|
1350 |
+
"base_uri": "https://localhost:8080/"
|
1351 |
+
},
|
1352 |
+
"id": "EGeLg-HXtMlu",
|
1353 |
+
"outputId": "c41250cd-b6e6-4b92-a775-d010fbdc803a"
|
1354 |
+
},
|
1355 |
+
"outputs": [
|
1356 |
+
{
|
1357 |
+
"name": "stdout",
|
1358 |
+
"output_type": "stream",
|
1359 |
+
"text": [
|
1360 |
+
"0.8875\n"
|
1361 |
+
]
|
1362 |
+
}
|
1363 |
+
],
|
1364 |
+
"source": [
|
1365 |
+
"print(sum(final_accuracies) / len(final_accuracies))"
|
1366 |
+
]
|
1367 |
+
}
|
1368 |
+
],
|
1369 |
+
"metadata": {
|
1370 |
+
"colab": {
|
1371 |
+
"provenance": []
|
1372 |
+
},
|
1373 |
+
"kernelspec": {
|
1374 |
+
"display_name": "Python 3",
|
1375 |
+
"name": "python3"
|
1376 |
+
},
|
1377 |
+
"language_info": {
|
1378 |
+
"name": "python"
|
1379 |
+
}
|
1380 |
+
},
|
1381 |
+
"nbformat": 4,
|
1382 |
+
"nbformat_minor": 0
|
1383 |
+
}
|
vercel.json
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"rewrites": [
|
3 |
+
{ "source": "/(.*)", "destination": "/api/app" }
|
4 |
+
]
|
5 |
+
}
|
weights.best.from_scratch1 (1).hdf5
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f84bea0167801a32295fa5321735e25b05f9b69b4c440dcdfe5f48da0db08b61
|
3 |
+
size 10380808
|