-
Notifications
You must be signed in to change notification settings - Fork 1
/
test_college.py
166 lines (150 loc) · 6.08 KB
/
test_college.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
import numpy as np
from sklearn.utils import shuffle
import math
def load_data_college(loc='data/local/CollegeOldData_HighAgreementPartialScoring.txt', skip_first_line=True):
allA, allB, allS = [],[],[]
with open(loc, 'rb') as f:
for line in f:
text = line.strip().split('\t')
allA.append(text[1])
allB.append(text[2])
allS.append(text[3])
if skip_first_line:
allA = allA[1:]
allB = allB[1:]
allS = allS[1:]
allS = [float(s) for s in allS]
## remove useless datapoints
index = [i for i, j in enumerate(allB) if (j == "empty" or ("I don't" in j))]
print("No. of empty and 'i don't know' cases': " , len(index))
index = [i for i, j in enumerate(allB) if (j == "empty" or ("I don't" in j) or ("\n" in j) or ('\"' in j) )]
print("No. of empty and 'i don't know' , 'i don't' and multi-line (suspicious) cases': " , len(index))
allA = np.asarray([x for i, x in enumerate(allA) if i not in index])
allB = np.asarray([x for i, x in enumerate(allB) if i not in index])
allS = np.asarray([x for i, x in enumerate(allS) if i not in index])
print("Average length of sentenceA ", sum(map(len, allA))/float(len(allA)))
print("Average length of sentenceB ", sum(map(len, allB))/float(len(allB)))
print len(allA), len(allB), len(allS)
## shuffle the data
allS, allA, allB = shuffle(allS, allA, allB, random_state=12345)
## split into 45% train, 5% dev and remaining ~50% test
trainA, devA, testA = allA[0 : int(math.floor(0.45 * len(allA)))], allA[int(math.floor(0.45 * len(allA))) + 1 : int(math.floor(0.5 * len(allA))) ], allA[int(math.floor(0.5 * len(allA))) + 1 : ]
trainB, devB, testB = allB[0 : int(math.floor(0.45 * len(allB)))], allB[int(math.floor(0.45 * len(allB))) + 1 : int(math.floor(0.5 * len(allB))) ], allB[int(math.floor(0.5 * len(allB))) + 1 : ]
trainS, devS, testS = allS[0 : int(math.floor(0.45 * len(allS)))], allS[int(math.floor(0.45 * len(allS))) + 1 : int(math.floor(0.5 * len(allS))) ], allS[int(math.floor(0.5 * len(allS))) + 1 : ]
print len(allA)
print len(trainA)+len(devA)+len(testA)
print len(trainA), len(devA), len(testA)
return [trainA, trainB, trainS], [devA, devB, devS], [testA, testB, testS]
def clip_data(data, N):
data[0] = data[0][0:N]
data[1] = data[1][0:N]
data[2] = data[2][0:N]
train, dev, test = load_data_college();
from encoders.classifier import Classifier;
###################################################################################################
c = Classifier(['feature_based']);
print "Training (feature_based)..."
c.classifier = c.train(train, dev)
c.classifier.save('pretrained/classifiers/fb-college.h5')
print "Trained performance (fb): "
c.test(test)
# ************ SUMMARY ***********
# Test data size: 2377
# Test Pearson: 0.780165116222
# Test Spearman: 0.773093078166
# Test MSE: 0.0618294011138
# ********************************
###################################################################################################
c = Classifier(['bow']);
# print "Untrained performance: "
# c.test(test)
# ************ SUMMARY ***********
# Test data size: 2366
# Test Pearson: 0.609879183339
# Test Spearman: 0.647898161358
# Test MSE: 0.146632924546
# ********************************
print "Training (bow)..."
c.classifier = c.train(train, dev)
c.classifier.save('pretrained/classifiers/bow-college.h5')
print "Trained performance (bow): "
c.test(test)
# ************ SUMMARY ***********
# Test data size: 2371
# Test Pearson: 0.815727992469
# Test Spearman: 0.814669747831
# Test MSE: 0.0529697648042
# ********************************
###################################################################################################
c = Classifier(['infersent']);
# N = 2000
# # clip_data(train, N)
# # clip_data(dev, N/2)
# # clip_data(test, N)
# print "Untrained performance: "
# c.test(test)
# # ************ SUMMARY ***********
# # Test data size: 2377
# # Test Pearson: 0.614370830801
# # Test Spearman: 0.704158202389
# # Test MSE: 0.142912132193
# # ********************************
print "Training (infersent)..."
c.classifier = c.train(train, dev)
c.classifier.save('pretrained/classifiers/infersent-college.h5')
print "Trained performance (infersent): "
c.test(test)
# ************ SUMMARY ***********
# Test data size: 2377
# Test Pearson: 0.864566020796
# Test Spearman: 0.85160629806
# Test MSE: 0.0403718484387
# ********************************
# Results with infersent trained on SICK data:
# (actually worse than untrained!)
# ************ SUMMARY ***********
# Test data size: 2377
# Test Pearson: 0.575628460689
# Test Spearman: 0.590724986585
# Test MSE: 0.155956669515
# ********************************
# ************ SUMMARY *********** with 0 score for contradiction cases
# Test data size: 2377
# Test Pearson: 0.425360783695
# Test Spearman: 0.445096454362
# Test MSE: 0.134747540396
# ********************************
###################################################################################################
c = Classifier(['bow', 'feature_based']);
print "Training (bow + feature_based)..."
c.classifier = c.train(train, dev)
c.classifier.save('pretrained/classifiers/bow_fb-college.h5');
print "Trained performance (bow + feature_based): "
c.test(test)
# ************ SUMMARY ***********
# Test data size: 2371
# Test Pearson: 0.851198922554
# Test Spearman: 0.838510775468
# Test MSE: 0.0441557772239
# ********************************
#
# # Results with bow+fb trained on SICK data:
# ************ SUMMARY ***********
# Test data size: 2366
# Test Pearson: 0.455408578396
# Test Spearman: 0.409406750153
# Test MSE: 0.142136707951
# ********************************
###################################################################################################
c = Classifier(['infersent', 'feature_based']);
print "Training (infersent + feature_based)..."
c.classifier = c.train(train, dev)
c.classifier.save('pretrained/classifiers/infersent_fb-college.h5');
print "Trained performance (infersent + feature based): "
c.test(test)
# ************ SUMMARY ***********
# Test data size: 2377
# Test Pearson: 0.886186401746
# Test Spearman: 0.868906840594
# Test MSE: 0.0339921869165
# ********************************