Spaces:
Sleeping
Sleeping
Upload 5 files
Browse files- app.py +20 -0
- cv.pkl +3 -0
- helper.py +266 -0
- model.pkl +3 -0
- requirements.txt +0 -0
app.py
ADDED
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import helper
|
3 |
+
import pickle
|
4 |
+
import sklearn
|
5 |
+
from sklearn.feature_extraction.text import CountVectorizer
|
6 |
+
|
7 |
+
model = pickle.load(open('model.pkl', 'rb'))
|
8 |
+
st.header("Duplicate Question Pairs")
|
9 |
+
|
10 |
+
q1 = st.text_input("Enter question 1")
|
11 |
+
q2 = st.text_input("Enter question 2")
|
12 |
+
|
13 |
+
if st.button("Find"):
|
14 |
+
query = helper.query_point_creator(q1, q2)
|
15 |
+
result = model.predict(query)[0]
|
16 |
+
|
17 |
+
if result:
|
18 |
+
st.header("Duplicate")
|
19 |
+
else:
|
20 |
+
st.header("Not Duplicate")
|
cv.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7990f044f2698296a6d944f70f0d10f55296eeb9544011d474748428deb2692c
|
3 |
+
size 543885
|
helper.py
ADDED
@@ -0,0 +1,266 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import pandas as pd
|
2 |
+
import numpy as np
|
3 |
+
from bs4 import BeautifulSoup
|
4 |
+
import re
|
5 |
+
|
6 |
+
import nltk
|
7 |
+
from nltk.corpus import stopwords
|
8 |
+
nltk.download('stopwords')
|
9 |
+
|
10 |
+
import distance
|
11 |
+
from fuzzywuzzy import fuzz
|
12 |
+
import pickle
|
13 |
+
|
14 |
+
import sklearn
|
15 |
+
from sklearn.feature_extraction.text import CountVectorizer
|
16 |
+
|
17 |
+
cv = pickle.load(open('cv.pkl', 'rb'))
|
18 |
+
def preprocess(q):
|
19 |
+
q = str(q).lower().strip()
|
20 |
+
|
21 |
+
q = q.replace('%', ' percent')
|
22 |
+
q = q.replace('$', ' dollar ')
|
23 |
+
q = q.replace('₹', ' rupee ')
|
24 |
+
q = q.replace('€', ' euro ')
|
25 |
+
q = q.replace('@', ' at ')
|
26 |
+
|
27 |
+
contractions = {
|
28 |
+
"ain't": "am not",
|
29 |
+
"aren't": "are not",
|
30 |
+
"can't": "can not",
|
31 |
+
"can't've": "can not have",
|
32 |
+
"'cause": "because",
|
33 |
+
"could've": "could have",
|
34 |
+
"couldn't": "could not",
|
35 |
+
"couldn't've": "could not have",
|
36 |
+
"didn't": "did not",
|
37 |
+
"doesn't": "does not",
|
38 |
+
"don't": "do not",
|
39 |
+
"hadn't": "had not",
|
40 |
+
"hadn't've": "had not have",
|
41 |
+
"hasn't": "has not",
|
42 |
+
"haven't": "have not",
|
43 |
+
"he'd": "he would",
|
44 |
+
"he'd've": "he would have",
|
45 |
+
"he'll": "he will",
|
46 |
+
"he'll've": "he will have",
|
47 |
+
"he's": "he is",
|
48 |
+
"how'd": "how did",
|
49 |
+
"how'd'y": "how do you",
|
50 |
+
"how'll": "how will",
|
51 |
+
"how's": "how is",
|
52 |
+
"i'd": "i would",
|
53 |
+
"i'd've": "i would have",
|
54 |
+
"i'll": "i will",
|
55 |
+
"i'll've": "i will have",
|
56 |
+
"i'm": "i am",
|
57 |
+
"i've": "i have",
|
58 |
+
"isn't": "is not",
|
59 |
+
"it'd": "it would",
|
60 |
+
"it'd've": "it would have",
|
61 |
+
"it'll": "it will",
|
62 |
+
"it'll've": "it will have",
|
63 |
+
"it's": "it is",
|
64 |
+
"let's": "let us",
|
65 |
+
"ma'am": "madam",
|
66 |
+
"mayn't": "may not",
|
67 |
+
"might've": "might have",
|
68 |
+
"mightn't": "might not",
|
69 |
+
"mightn't've": "might not have",
|
70 |
+
"must've": "must have",
|
71 |
+
"mustn't": "must not",
|
72 |
+
"mustn't've": "must not have",
|
73 |
+
"needn't": "need not",
|
74 |
+
"needn't've": "need not have",
|
75 |
+
"o'clock": "of the clock",
|
76 |
+
"oughtn't": "ought not",
|
77 |
+
"oughtn't've": "ought not have",
|
78 |
+
"shan't": "shall not",
|
79 |
+
"sha'n't": "shall not",
|
80 |
+
"shan't've": "shall not have",
|
81 |
+
"she'd": "she would",
|
82 |
+
"she'd've": "she would have",
|
83 |
+
"she'll": "she will",
|
84 |
+
"she'll've": "she will have",
|
85 |
+
"she's": "she is",
|
86 |
+
"should've": "should have",
|
87 |
+
"shouldn't": "should not",
|
88 |
+
"shouldn't've": "should not have",
|
89 |
+
"so've": "so have",
|
90 |
+
"so's": "so as",
|
91 |
+
"that'd": "that would",
|
92 |
+
"that'd've": "that would have",
|
93 |
+
"that's": "that is",
|
94 |
+
"there'd": "there would",
|
95 |
+
"there'd've": "there would have",
|
96 |
+
"there's": "there is",
|
97 |
+
"they'd": "they would",
|
98 |
+
"they'd've": "they would have",
|
99 |
+
"they'll": "they will",
|
100 |
+
"they'll've": "they will have",
|
101 |
+
"they're": "they are",
|
102 |
+
"they've": "they have",
|
103 |
+
"to've": "to have",
|
104 |
+
"wasn't": "was not",
|
105 |
+
"we'd": "we would",
|
106 |
+
"we'd've": "we would have",
|
107 |
+
"we'll": "we will",
|
108 |
+
"we'll've": "we will have",
|
109 |
+
"we're": "we are",
|
110 |
+
"we've": "we have",
|
111 |
+
"weren't": "were not",
|
112 |
+
"what'll": "what will",
|
113 |
+
"what'll've": "what will have",
|
114 |
+
"what're": "what are",
|
115 |
+
"what's": "what is",
|
116 |
+
"what've": "what have",
|
117 |
+
"when's": "when is",
|
118 |
+
"when've": "when have",
|
119 |
+
"where'd": "where did",
|
120 |
+
"where's": "where is",
|
121 |
+
"where've": "where have",
|
122 |
+
"who'll": "who will",
|
123 |
+
"who'll've": "who will have",
|
124 |
+
"who's": "who is",
|
125 |
+
"who've": "who have",
|
126 |
+
"why's": "why is",
|
127 |
+
"why've": "why have",
|
128 |
+
"will've": "will have",
|
129 |
+
"won't": "will not",
|
130 |
+
"won't've": "will not have",
|
131 |
+
"would've": "would have",
|
132 |
+
"wouldn't": "would not",
|
133 |
+
"wouldn't've": "would not have",
|
134 |
+
"y'all": "you all",
|
135 |
+
"y'all'd": "you all would",
|
136 |
+
"y'all'd've": "you all would have",
|
137 |
+
"y'all're": "you all are",
|
138 |
+
"y'all've": "you all have",
|
139 |
+
"you'd": "you would",
|
140 |
+
"you'd've": "you would have",
|
141 |
+
"you'll": "you will",
|
142 |
+
"you'll've": "you will have",
|
143 |
+
"you're": "you are",
|
144 |
+
"you've": "you have"
|
145 |
+
}
|
146 |
+
|
147 |
+
q_decontracted = []
|
148 |
+
|
149 |
+
for word in q.split():
|
150 |
+
if word in contractions:
|
151 |
+
word = contractions[word]
|
152 |
+
|
153 |
+
q_decontracted.append(word)
|
154 |
+
|
155 |
+
q = ' '.join(q_decontracted)
|
156 |
+
q = q.replace("'ve", " have")
|
157 |
+
q = q.replace("n't", " not")
|
158 |
+
q = q.replace("'re", " are")
|
159 |
+
q = q.replace("'ll", " will")
|
160 |
+
|
161 |
+
# Removing HTML tags
|
162 |
+
q = BeautifulSoup(q)
|
163 |
+
q = q.get_text()
|
164 |
+
|
165 |
+
# Remove punctuations
|
166 |
+
pattern = re.compile('\W')
|
167 |
+
q = re.sub(pattern, ' ', q).strip()
|
168 |
+
|
169 |
+
return q
|
170 |
+
|
171 |
+
|
172 |
+
def test_common_words(q1, q2):
|
173 |
+
w1 = set(map(lambda words : words.lower().strip(), q1.split(" ")))
|
174 |
+
w2 = set(map(lambda words : words.lower().strip(), q2.split(" ")))
|
175 |
+
return len(w1 & w2)
|
176 |
+
|
177 |
+
def test_total_words(q1, q2):
|
178 |
+
w1 = set(map(lambda words : words.strip(), q1.split(" ")))
|
179 |
+
w2 = set(map(lambda words : words.strip(), q2.split(" ")))
|
180 |
+
return (len(w1) + len(w2))
|
181 |
+
|
182 |
+
def fetch_test_features(q1, q2):
|
183 |
+
SAFE_DIV = 0.00001
|
184 |
+
STOP_WORDS = stopwords.words('english')
|
185 |
+
token_features = [0.0] * 11
|
186 |
+
|
187 |
+
q1_tokens = q1.split()
|
188 |
+
q2_tokens = q2.split()
|
189 |
+
|
190 |
+
if len(q1_tokens) == 0 or len(q2_tokens) == 0:
|
191 |
+
return token_features
|
192 |
+
|
193 |
+
q1_words = set([word for word in q1_tokens if word not in STOP_WORDS])
|
194 |
+
q2_words = set([word for word in q2_tokens if word not in STOP_WORDS])
|
195 |
+
|
196 |
+
q1_stop = set([word for word in q1_tokens if word in STOP_WORDS])
|
197 |
+
q2_stop = set([word for word in q2_tokens if word in STOP_WORDS])
|
198 |
+
|
199 |
+
common_word_cnt = len(q1_words.intersection(q2_words))
|
200 |
+
common_stop_cnt = len(q1_stop.intersection(q2_stop))
|
201 |
+
common_token_cnt = len(set(q1_tokens).intersection(set(q2_tokens)))
|
202 |
+
|
203 |
+
token_features[0] = round(common_word_cnt / (min(len(q1_words), len(q2_words)) + SAFE_DIV), 2)
|
204 |
+
token_features[1] = round(common_word_cnt / (max(len(q1_words), len(q2_words)) + SAFE_DIV), 2)
|
205 |
+
|
206 |
+
token_features[2] = round(common_stop_cnt / (min(len(q1_stop), len(q2_stop)) + SAFE_DIV), 2)
|
207 |
+
token_features[3] = round(common_stop_cnt / (max(len(q1_stop), len(q2_stop)) + SAFE_DIV), 2)
|
208 |
+
|
209 |
+
token_features[4] = round(common_token_cnt / (min(len(q1_tokens), len(q2_tokens)) + SAFE_DIV), 2)
|
210 |
+
token_features[5] = round(common_token_cnt / (max(len(q1_tokens), len(q2_tokens)) + SAFE_DIV), 2)
|
211 |
+
|
212 |
+
token_features[6] = int(q1_tokens[-1] == q2_tokens[-1])
|
213 |
+
token_features[7] = int(q1_tokens[0] == q2_tokens[0])
|
214 |
+
|
215 |
+
token_features[8] = abs(len(q1_tokens) - len(q2_tokens))
|
216 |
+
token_features[9] = round((len(q1_tokens) - len(q2_tokens)) / 2, 2)
|
217 |
+
|
218 |
+
strs = list(distance.lcsubstrings(q1, q2))
|
219 |
+
token_features[10] = round(len(strs[0]) / (min(len(q1), len(q2)) + 1), 2)
|
220 |
+
return token_features
|
221 |
+
|
222 |
+
|
223 |
+
def fetch_test_fuzzy_features(q1, q2):
|
224 |
+
fuzzy_features = [0.0] * 4
|
225 |
+
|
226 |
+
# fuzz_ratio
|
227 |
+
fuzzy_features[0] = fuzz.QRatio(q1, q2)
|
228 |
+
|
229 |
+
# fuzz_partial_ratio
|
230 |
+
fuzzy_features[1] = fuzz.partial_ratio(q1, q2)
|
231 |
+
|
232 |
+
# token_sort_ratio
|
233 |
+
fuzzy_features[2] = fuzz.token_sort_ratio(q1, q2)
|
234 |
+
|
235 |
+
# token_set_ratio
|
236 |
+
fuzzy_features[3] = fuzz.token_set_ratio(q1, q2)
|
237 |
+
|
238 |
+
return fuzzy_features
|
239 |
+
|
240 |
+
|
241 |
+
def query_point_creator(q1, q2):
|
242 |
+
input_query = []
|
243 |
+
q1 = preprocess(q1)
|
244 |
+
q2 = preprocess(q2)
|
245 |
+
|
246 |
+
input_query.append(len(q1))
|
247 |
+
input_query.append(len(q2))
|
248 |
+
|
249 |
+
input_query.append(len(q1.split(" ")))
|
250 |
+
input_query.append(len(q2.split(" ")))
|
251 |
+
|
252 |
+
input_query.append(test_common_words(q1, q2))
|
253 |
+
input_query.append(test_total_words(q1, q2))
|
254 |
+
|
255 |
+
input_query.append(round(input_query[4] / input_query[5], 2))
|
256 |
+
|
257 |
+
token_features = fetch_test_features(q1, q2)
|
258 |
+
input_query.extend(token_features)
|
259 |
+
|
260 |
+
fuzzy_features = fetch_test_fuzzy_features(q1, q2)
|
261 |
+
input_query.extend(fuzzy_features)
|
262 |
+
|
263 |
+
q1_arr = cv.transform([q1]).toarray()
|
264 |
+
q2_arr = cv.transform([q2]).toarray()
|
265 |
+
|
266 |
+
return np.hstack((np.array(input_query).reshape(1, 22), q1_arr, q2_arr))
|
model.pkl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3e09f34acdafdfb0dbe60ee9c92b82478dfdb3316dc5acfc4268f067985063b5
|
3 |
+
size 88637720
|
requirements.txt
ADDED
Binary file (172 Bytes). View file
|
|