-
Notifications
You must be signed in to change notification settings - Fork 5
/
sensitivity_tests.py
170 lines (153 loc) · 8.63 KB
/
sensitivity_tests.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
import re
import os
import argparse
import random
import copy
import numpy as np
from proc_datasets import *
from io import open
def convert_to_experiment_grouping(datadict):
conddict = {}
for k in datadict:
it = datadict[k]['item']
co = datadict[k]['cond']
if it not in conddict:
conddict[it] = {}
conddict[it][co] = {}
for keycat in ['sent','tgt','tgtcloze','constraint','licensing','maxcloze','tgtprob','toppreds','topprobs']:
if keycat in datadict[k]:
conddict[it][co][keycat] = datadict[k][keycat]
return conddict
def cprag_sensitivity_test(dataset_ref,target_probs):
for i,prob in enumerate(target_probs):
dataset_ref[i]['tgtprob'] = prob
conddict = convert_to_experiment_grouping(dataset_ref)
thresh = 0.01
exp_top = {'H':[],'L':[]}
exp_top_thresh = {'H':[],'L':[]}
allprobs = {'H':[],'L':[]}
for it in conddict:
exp_prob,wc_prob,bc_prob = [conddict[it][cont]['tgtprob'] for cont in ['exp','wc','bc']]
cons = conddict[it]['exp']['constraint']
allprobs[cons].append((exp_prob,wc_prob,bc_prob))
if (exp_prob > wc_prob) and (exp_prob > bc_prob):
exp_top[cons].append(1)
if abs(exp_prob - wc_prob) > thresh and abs(exp_prob - bc_prob) > thresh:
exp_top_thresh[cons].append(1)
else:
exp_top_thresh[cons].append(0)
else:
exp_top[cons].append(0)
exp_top_thresh[cons].append(0)
report = '\nSensitivity results:\n\n'
report += 'Expected word more probable than two inappropriate targets: %s (%s/%s)\n'%(get_acc(exp_top['H']+exp_top['L']),sum(exp_top['H']+exp_top['L']),len(exp_top['H']+exp_top['L']))
report += ' for high-constraint items: %s (%s/%s)\n'%(get_acc(exp_top['H']),sum(exp_top['H']),len(exp_top['H']))
report += ' for low-constraint items: %s (%s/%s)\n\n'%(get_acc(exp_top['L']),sum(exp_top['L']),len(exp_top['L']))
report += 'Expected word more probable than two inappropriate targets -- difference threshold %s: %s (%s/%s)\n'%(thresh,get_acc(exp_top_thresh['H']+exp_top_thresh['L']),sum(exp_top_thresh['H']+exp_top_thresh['L']),len(exp_top_thresh['H']+exp_top_thresh['L']))
report += ' for high-constraint items: %s (%s/%s)\n'%(get_acc(exp_top_thresh['H']),sum(exp_top_thresh['H']),len(exp_top_thresh['H']))
report += ' for low-constraint items: %s (%s/%s)\n'%(get_acc(exp_top_thresh['L']),sum(exp_top_thresh['L']),len(exp_top_thresh['L']))
return report
def role_sensitivity_test(dataset_ref,target_probs):
dataset_dict,clozelist = dataset_ref
for i,prob in enumerate(target_probs):
dataset_dict[i]['tgtprob'] = prob
conddict = convert_to_experiment_grouping(dataset_dict)
thresh = 0.01
good_top = []
good_top_thresh = []
probpairs = []
clozepairs = []
for it in conddict:
a_prob,b_prob = (conddict[it]['a']['tgtprob'],conddict[it]['b']['tgtprob'])
if (a_prob > b_prob):
good_top.append(1)
else:
good_top.append(0)
if (a_prob > b_prob) and (abs(a_prob - b_prob) > thresh):
good_top_thresh.append(1)
else:
good_top_thresh.append(0)
probpairs.append((a_prob,b_prob))
clozepairs.append((conddict[it]['a']['tgtcloze'],conddict[it]['b']['tgtcloze']))
probdiffs = [e[0] - e[1] for e in probpairs]
clozediffs = [e[0] - e[1] for e in clozepairs]
report = '\nSensitivity results:\n\n'
report += 'Good completion more probable than role reversal: %s (%s/%s)\n\n'%(get_acc(good_top),sum(good_top),len(good_top))
report += 'Good completion more probable than role reversal -- difference threshold %s: %s (%s/%s)\n\n'%(thresh,get_acc(good_top_thresh),sum(good_top_thresh),len(good_top_thresh))
report += 'AVG PROB DIFF: %s\n'%np.average(probdiffs)
report += 'AVG CLOZE DIFF: %s\n'%np.average(clozediffs)
return report
def neg_sensitivity_test(dataset_ref,target_probs):
for i,prob in enumerate(target_probs):
dataset_ref[i]['tgtprob'] = prob
conddict = convert_to_experiment_grouping(dataset_ref)
thresh = 0.01
pattern = []
same = []
preftrue = {'aff':[],'neg':[]}
preftrue_l = {'aff':[],'neg':[]}
preftrue_u = {'aff':[],'neg':[]}
preftrue_thresh = {'aff':[],'neg':[]}
lic = None
for it in conddict:
if 'licensing' in conddict[it]['TA']:
lic = conddict[it]['TA']['licensing']
for true_cond,false_cond,pol in [('TA','FA','aff'),('TN','FN','neg')]:
true_prob,false_prob = (conddict[it][true_cond]['tgtprob'],conddict[it][false_cond]['tgtprob'])
if true_prob > false_prob:
score = 1
else:
score = 0
preftrue[pol].append(score)
if lic:
if lic == 'Y':
preftrue_l[pol].append(score)
elif lic == 'N':
preftrue_u[pol].append(score)
if (true_prob > false_prob) and (abs(true_prob - false_prob) > thresh):
preftrue_thresh[pol].append(1)
else:
preftrue_thresh[pol].append(0)
report = '\nSensitivity results:\n\n'
report += 'True completion more probable than false: %s (%s/%s)\n'%(get_acc(preftrue['aff'] + preftrue['neg']),sum(preftrue['aff'] + preftrue['neg']),len(preftrue['aff'] + preftrue['neg']))
report += ' in affirmative contexts: %s (%s/%s)\n'%(get_acc(preftrue['aff']),sum(preftrue['aff']),len(preftrue['aff']))
report += ' in negative contexts: %s (%s/%s)\n'%(get_acc(preftrue['neg']),sum(preftrue['neg']),len(preftrue['neg']))
report += 'True completion more probable than false -- difference threshold %s: %s (%s/%s)\n'%(thresh,get_acc(preftrue_thresh['aff']+preftrue_thresh['neg']),sum(preftrue_thresh['aff']+preftrue_thresh['neg']),len(preftrue_thresh['aff']+preftrue_thresh['neg']))
report += ' in affirmative contexts: %s (%s/%s)\n'%(get_acc(preftrue_thresh['aff']),sum(preftrue_thresh['aff']),len(preftrue_thresh['aff']))
report += ' in negative contexts: %s (%s/%s)\n\n'%(get_acc(preftrue_thresh['neg']),sum(preftrue_thresh['neg']),len(preftrue_thresh['neg']))
if lic:
report += 'True completion more probable in NATURAL sentences: %s (%s/%s)\n'%(get_acc(preftrue_l['aff'] + preftrue_l['neg']),sum(preftrue_l['aff'] + preftrue_l['neg']),len(preftrue_l['aff'] + preftrue_l['neg']))
report += ' in affirmative contexts: %s (%s/%s)\n'%(get_acc(preftrue_l['aff']),sum(preftrue_l['aff']),len(preftrue_l['aff']))
report += ' in negative contexts: %s (%s/%s)\n'%(get_acc(preftrue_l['neg']),sum(preftrue_l['neg']),len(preftrue_l['neg']))
report += 'True completion more probable in LESS NATURAL sentences: %s (%s/%s)\n'%(get_acc(preftrue_u['aff'] + preftrue_u['neg']),sum(preftrue_u['aff'] + preftrue_u['neg']),len(preftrue_u['aff'] + preftrue_u['neg']))
report += ' in affirmative contexts: %s (%s/%s)\n'%(get_acc(preftrue_u['aff']),sum(preftrue_u['aff']),len(preftrue_u['aff']))
report += ' in negative contexts: %s (%s/%s)\n'%(get_acc(preftrue_u['neg']),sum(preftrue_u['neg']),len(preftrue_u['neg']))
report += '\n\n'
return report
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--probdir",default=None, type=str)
parser.add_argument("--resultsdir",default=None, type=str)
parser.add_argument("--models", nargs="+", type=str)
parser.add_argument("--cprag_stim", default=None, type=str)
parser.add_argument("--role_stim", default=None, type=str)
parser.add_argument("--negsimp_stim", default=None, type=str)
parser.add_argument("--negnat_stim", default=None, type=str)
args = parser.parse_args()
testlist = [
(args.cprag_stim, cprag_sensitivity_test,'cprag',process_cprag),
(args.role_stim, role_sensitivity_test,'role',process_role),
(args.negsimp_stim, neg_sensitivity_test,'negsimp',process_negsimp),
(args.negnat_stim, neg_sensitivity_test,'negnat',process_negnat)
]
for stimfile,sens_test,testname,process_func in testlist:
if not stimfile: continue
inputlist,_,dataset_ref = process_func(stimfile,mask_tok=False)
with open(args.resultsdir+'/sensitivity-%s.txt'%testname,'w') as out:
for modelname in args.models:
out.write('\n\n***\nMODEL: %s\n***\n'%modelname)
target_probs = []
with open(os.path.join(args.probdir,'modeltgtprobs-%s-%s'%(testname,modelname))) as probfile:
for line in probfile: target_probs.append(float(line.strip()))
report = sens_test(dataset_ref,target_probs)
out.write(report)