-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathfeature_extraction.py
165 lines (130 loc) · 4.8 KB
/
feature_extraction.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
from fuzzywuzzy import fuzz
import numpy as np
import nltk
from nltk import word_tokenize, sent_tokenize
wpt = nltk.WordPunctTokenizer()
def cosine(u, v):
return (np.dot(u, v) / (np.linalg.norm(u) * np.linalg.norm(v)))
#For baroni et al embeddings:- Word2Vec
def computeVecSum(vectors):
n = len(vectors)
d = 399
s = []
for i in range(d):
s.append(0)
s = np.array(s)
for vec in vectors:
s = s + np.array(vec)
return (s)
def cos_sim(df, text_colname, ans_colname, emb):
cos = []
for t in range(len(df)):
model_answer = word_tokenize(list(df[ans_colname])[t])
if type(df[text_colname][t]) == str:
texts = wpt.tokenize(list(df[text_colname])[t])
word_vecM = [emb[i] for i in model_answer if i in emb]
word_vecR = [emb[i] for i in texts if i in emb]
sent_vecM = computeVecSum(word_vecM)
sent_vecR = computeVecSum(word_vecR)
cos.append(cosine(sent_vecM,sent_vecR))
else:
cos.append(0)
return cos
def cos_wm(df, qn_colname, text_colname, ans_colname, score_colname, emb):
cos = []
for i in range(len(df)):
sim = []
sim_c = 0
model_answer = word_tokenize(list(df[ans_colname])[i])
word_vecM = [emb[k] for k in model_answer if k in emb]
sent_vecM = computeVecSum(word_vecM)
for j in range(len(df)):
if i == j:
continue
elif df[qn_colname][i] == df[qn_colname][j] and df[score_colname][j] == 5 and type(df[text_colname][j]) == str:
texts = wpt.tokenize(list(df[text_colname])[j])
word_vecR = [emb[k] for k in texts if k in emb]
sent_vecR = computeVecSum(word_vecR)
sim.append(cosine(sent_vecM,sent_vecR))
sim_c+=5
if type(df[text_colname][i]) == str:
texts = wpt.tokenize(list(df[text_colname])[i])
word_vecR = [emb[k] for k in texts if k in emb]
sent_vecR = computeVecSum(word_vecR)
c = (sim_c*cosine(sent_vecM,sent_vecR) + sum(sim))/(2*sim_c)
elif sim_c != 0:
c = sum(sim)/(2*sim_c)
else:
c = 0
cos.append(c)
return cos
def alignment(df, text_colname, ans_colname, emb):
align = []
for i in range(len(df)):
if type(df[text_colname][i]) == str:
test_model_answer = word_tokenize(list(df[ans_colname])[i])
test_text = word_tokenize(list(df[text_colname])[i])
x, y = [], []
for ma in test_model_answer:
for tt in test_text:
if ma in emb and tt in emb:
ma_embedding = emb[ma]
tt_embedding = emb[tt]
cos_similarity = cosine(ma_embedding, tt_embedding)
if cos_similarity >= 0.4:
x.append(ma)
y.append(tt)
alignment_score = (len(set(x)) + len(set(y))) / (len(set(test_model_answer)) + len(set(test_text)))
align.append(alignment_score)
else:
align.append(0)
return align
def length(df, text_colname, ans_colname):
length_ratio = []
for i in range(len(df)):
if type(df[text_colname][i]) == str:
test_model_answer = word_tokenize(list(df[ans_colname])[i])
test_text = word_tokenize(list(df[text_colname])[i])
lr = len(set(test_text)) / len(set(test_model_answer))
length_ratio.append(round(lr, 3))
else:
length_ratio.append(0)
return length_ratio
def eucledian(df, text_colname, ans_colname, emb):
eucledian_distance = []
for t in range(len(df)):
model_answer = word_tokenize(list(df[ans_colname])[t])
if type(df[text_colname][t]) == str:
texts = wpt.tokenize(list(df[text_colname])[t])
word_vecM = [emb[i] for i in model_answer if i in emb]
word_vecR = [emb[i] for i in texts if i in emb]
sent_vecM = computeVecSum(word_vecM)
sent_vecR = computeVecSum(word_vecR)
ed = round(np.linalg.norm(sent_vecM-sent_vecR), 3)
else:
ed=0
eucledian_distance.append(ed)
return eucledian_distance
def fuzzy_features(df, text_colname, ans_colname):
fuzzy_ratio = []
fuzzy_partial_ratio = []
fuzzy_token_sort = []
fuzzy_token_set = []
for t in range(len(df)):
model_answer = df[ans_colname][t]
if type(df[text_colname][t]) == str:
texts = df[text_colname][t]
fuzzy_r = fuzz.ratio(model_answer, texts)
fuzzy_pr = fuzz.partial_ratio(model_answer, texts)
fuzzy_tsort = fuzz.token_sort_ratio(model_answer, texts)
fuzzy_tset = fuzz.token_set_ratio(model_answer, texts)
else:
fuzzy_r = 0
fuzzy_pr = 0
fuzzy_tsort = 0
fuzzy_tset = 0
fuzzy_ratio.append(fuzzy_r / 100)
fuzzy_partial_ratio.append(fuzzy_pr / 100)
fuzzy_token_sort.append(fuzzy_tsort / 100)
fuzzy_token_set.append(fuzzy_tset / 100)
return [fuzzy_ratio, fuzzy_partial_ratio, fuzzy_token_sort, fuzzy_token_set]