-
Notifications
You must be signed in to change notification settings - Fork 0
/
alignment_utils.py
367 lines (291 loc) · 12.9 KB
/
alignment_utils.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
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
import ilp_utils
from word2vec import *
from ilp_utils import *
import json
import pandas as pd
import itertools
import numpy as np
from frame_extraction import *
from operator import itemgetter
import os
import math
from nltk.internals import find_jars_within_path
from nltk.tag import StanfordPOSTagger
import subprocess
#function to compute the sentences dataframe from the data.also includes ov1- index where argument and sentence overlap.This helps for finding neighbouring frames
#returns sentences dataframe
def extract_testsen(json_filename):
data=load_srl_data(json_filename)
sen_t1=pd.DataFrame(columns=['sen_id','Sentence','Process','arg_id','Arg','role'])
c=0
for p in data:
for i in data[p][1]:
if data[p][1][i]==[]:
#print "Breaking",i
continue
#print data[p]
for n in data[p][1][i]:
#print n
sen_t1.loc[c,'Process']=p
sen_t1.loc[c,'Arg']=n[1]
sen_t1.loc[c,'sen_id']=i
sen_t1.loc[c,'arg_id']=int(n[0])
id1=(i,int(n[0]))
sen_t1.loc[c,'role']=data[p][3][id1][0]
for d in data[p][0]:
if data[p][0][d]==i:
sen_t1.loc[c,'Sentence']=d
c=c+1
#creating replica of arguments so that they can be modified
sen_t1['Arg_dup']=sen_t1['Arg']
for i in list(sen_t1.index.get_values()):
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace('-LRB- ','(')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace(' -RRB-',')')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace(' ,',',')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace('-LSB- ','[')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace(' -RSB-',']')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace(' \'' , '\'')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace(' ;',';')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace(' .','.')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace(' - ','- ')
sen_t1.loc[i,'Arg_dup']=sen_t1.loc[i,'Arg_dup'].replace(' :',':')
sen_t1.loc[i,'o1']=sen_t1.loc[i,'Sentence'].find(sen_t1.loc[i,'Arg_dup'])
#sen_t1.to_csv('NEWSENDATAQA.csv',sep='\t')
return sen_t1
def extract_goldsen(json_filename):
d_gold = json.load(open(json_filename, "r"))
gold_data =get_gold_data(d_gold)
sentences=pd.DataFrame(columns=['sen_id','Sentence','Process','Arg','role'])
n=0
#getting sentence data
data=load_srl_data(json_filename)
s={}
for p in data:
s[p]=data[p][4]
for i in gold_data :
sentences.loc[n,'sen_id']=i[0]
sentences.loc[n,'Arg']=i[1]
sentences.loc[n,'Process']=i[4]
sentences.loc[n,'role']=gold_data[i]
ind=sentences.loc[n,'sen_id']
p=sentences.loc[n,'Process']
sentences.loc[n,'Sentence']=s[p][ind]
n=n+1
#creating replica of arguments so that they can be modified
sentences['Arg_dup']=sentences['Arg']
#optional:delete spaces near special characters
for i in list(sentences.index.get_values()):
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace('-LRB- ','(')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace(' -RRB-',')')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace(' ,',',')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace('-LSB- ','[')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace(' -RSB-',']')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace(' \'' , '\'')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace(' ;',';')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace(' .','.')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace(' - ','- ')
sentences.loc[i,'Arg_dup']=sentences.loc[i,'Arg_dup'].replace(' :',':')
#finding index of argument in sentence
sentences.loc[i,'o1']=sentences.loc[n,'Sentence'].find(sentences.loc[n,'Arg_dup'])
return sentences
def get_frame_feature(sen_t1):
sen_t1['Sentence'].to_csv('/home/sadhana/semafor-semantic-parser/file_2/sentences.txt')
subprocess.call(['/home/sadhana/semafor-semantic-parser/release/fnParserDriver.sh','/home/sadhana/semafor-semantic-parser/file_2/sentences.txt'])
frame_test=get_frames('sentences.txt.out')#Insert appropriate semafor output file
print "Getting sen features "
for i in list(sen_t1.index.get_values()):
fs=frame_test[i]
fs.sort(key=itemgetter(2),reverse=False)
maxi1=0
mini1=1000
le=int(sen_t1.loc[i,'o1'])
for j in range(0,len(fs)):
e=int(fs[j][2])
s=int(fs[j][1])
if(e > maxi1) and ( e < le):
lf=fs[j][0]
maxi1=int(fs[j][2])
else:
lf=''
if(s < mini1) and (s > (le +len(sen_t1.loc[i,'Arg']))):
mini1= int(fs[j][1])
rf=fs[j][0]
else:
rf=''
sen_t1.loc[i,'lf1']=lf
sen_t1.loc[i,'rf1']=rf
return sen_t1
def get_pos_tag(sen):
os.environ['CLASSPATH']='STANFORDTOOLSDIR/stanford-postagger-full-2015-12-09/stanford-postagger.jar' #set classpath to pos tagger
os.environ['STANFORD_MODELS']='STANFORDTOOLSDIR/stanford-postagger-full-2015-12-09/models'
st = StanfordPOSTagger('/home/sadhana/stanford-postagger-full-2015-12-09/models/english-left3words-distsim.tagger',path_to_jar=
'/home/sadhana/stanford-postagger-full-2015-12-09/stanford-postagger.jar')#,path_to_models_jar='/home/sadhana/stanford-postagger-full-2015-12-09/models')
stanford_dir = st._stanford_jar.rpartition('/')[0]
stanford_jars = find_jars_within_path(stanford_dir)
st._stanford_jar = ':'.join(stanford_jars)
for i in list(sen.index.get_values()):
t=st.tag(sen.loc[i,'Arg'].split())
tags=[]
for j in range(0,len(t)):
tags.append(t[j][1])
#print i
sen.set_value(i,'POStag',tags)
return sen
def jaccard_similarity(x,y):
intersection_cardinality = len(set.intersection(*[set(x), set(y)]))
union_cardinality = len(set.union(*[set(x), set(y)]))
return intersection_cardinality/float(union_cardinality)
def cosine_similarity(x,y):
cs = np.dot(x,y)/(np.linalg.norm(x)* np.linalg.norm(y))
score=(cs+1)/2
return score
def get_rf_cosine_similarity(df):
w=Word2VecModel()
for j in list(df.index.get_values()):
try:
if type(df.loc[j,'rf1'])==float:
if math.isnan(df.loc[j,'rf1']) :
v1=[]
else:
df.loc[j,'rf1']=df.loc[j,'rf1'].replace('_',' ')
v1=w.get_sent_vector(df.loc[j,'rf1'])
if type(df.loc[j,'rf2'])==float:
if math.isnan(df.loc[j,'rf2']):
v2=[]
else:
df.loc[j,'rf2']=df.loc[j,'rf2'].replace('_',' ')
v2=w.get_sent_vector(df.loc[j,'rf2'])
if v1!=[] and v2!=[]:
score=cosine_similarity(v1,v2)
else:
score=0
except ValueError:
score=0
df.loc[j,'rf_Cscore']=score
return df
def get_lf_cosine_similarity(df):
w=Word2VecModel()
for j in list(df.index.get_values()):
try:
if type(df.loc[j,'lf1'])==float:
if math.isnan(df.loc[j,'lf1']) :
v1=[]
else:
df.loc[j,'lf1']=df.loc[j,'lf1'].replace('_',' ')
v1=w.get_sent_vector(df.loc[j,'lf1'])
if type(df.loc[j,'lf2'])==float:
if math.isnan(df.loc[j,'lf2']):
v2=[]
else:
df.loc[j,'lf2']=df.loc[j,'lf2'].replace('_',' ')
v2=w.get_sent_vector(df.loc[j,'lf2'])
if v1!=[] and v2!=[]:
score=cosine_similarity(v1,v2)
else:
score=0
except ValueError:
score=0
df.loc[j,'lf_Cscore']=score
return df
def get_arg_cosine_simialrity(df):
w=Word2VecModel()
for j in list(df.index.get_values()):
try:
v1=w.get_sent_vector(df.loc[j,'Arg1'])
v2=w.get_sent_vector(df.loc[j,'Arg2'])
score=cosine_similarity(v1,v2)
except ValueError:
score=np.nan
df.loc[j,'Cscore']=score
return df
def get_entailment_score(df):
for j in list(df.index.get_values()):
df.loc[j,'Escore']=get_similarity_score(df.loc[j,'Arg1'],df.loc[j,'Arg2'])
return df
def get_pos_similarity(df):
for i in list(df.index.get_values()):
df.loc[i,'POSsim']=jaccard_similarity(df.loc[i,'POStag1'],df.loc[i,'POStag2'])
def get_arg_pairs(sentences):
df=pd.DataFrame(columns=['Process','Arg1','Arg2','Sentence1','Sentence2','Role1','Role2','True_label'])
m=0
processes=list(set(sentences['Process']))
for p in processes:
for ac in itertools.combinations(sentences[sentences['Process']==p].index.tolist(),2):
df.loc[m,'Process']=p
df.loc[m,'Arg1']=sentences.loc[ac[0],'Arg']
df.loc[m,'Arg2']=sentences.loc[ac[1],'Arg']
df.loc[m,'Sentence1']=sentences.loc[ac[0],'Sentence']
df.loc[m,'Sentence2']=sentences.loc[ac[1],'Sentence']
df.loc[m,'Role1']=sentences.loc[ac[0],'Role']
df.loc[m,'Role2']=sentences.loc[ac[1],'Role']
if df.loc[m,'Role1']==df.loc[m,'Role2']:
df.loc[m,'True_label']=1
else:
df.loc[m,'True_label']=0
m=m+1
return df
def merge_sen_df(sentences,df):
for i in list(df.index.get_values()):
print i
arg=df.loc[i,'Arg1']
y=sentences[(sentences['Arg']==arg) & (sentences['Process']==df.loc[i,'Process']) ]
z=sentences[(sentences['Arg']==df.loc[i,'Arg2']) & (sentences['Process']==df.loc[i,'Process']) ]
df.loc[i,'Sentence1']=y['Sentence'].values[0]
df.loc[i,'Sentence2']=z['Sentence'].values[0]
df.loc[i,'lf1']=y['lf1'].values[0]
df.loc[i,'rf1']=y['rf1'].values[0]
df.loc[i,'lf2']=z['lf1'].values[0]
df.loc[i,'rf2']=z['rf1'].values[0]
df.loc[i,'o1']=y['o1'].values[0]
df.loc[i,'o2']=z['o1'].values[0]
df.set_value(i,'POStag1',y['POStag'].values[0])
df.set_value(i,'POStag2',z['POStag'].values[0])
return df
def plot_precision_yield(plot_data):
srl_plot_df = plot_data
srl_plot_df = srl_plot_df.iloc[10:]
# plot size
plt.rc('figure', figsize=(18,12))
# plot lines
plt.plot(srl_plot_df.index, srl_plot_df.precision, label=r'POS', linewidth=3)
# configure plot
plt.tick_params(axis='both', which='major', labelsize=50)
plt.xlabel('Recall', fontsize=50)
plt.ylabel('Precison', fontsize=50)
plt.xlim([0, 1])
plt.ylim([0, 1.005])
plt.legend(loc='lower right', handlelength=3, prop={'size':45}) #borderpad=1.5, labelspacing=1.5,
plt.tight_layout()
plt.show()
def plot_pr_overall_concatenated(srl_all_data):
"""Plots overall precision recall after joining data from all the folds"""
# concatenate data from all folds into a single dictionary which has
# as key (sentence_id, start_index, end_index)
# as value (gold_role, (predicted_role, prediction_score))
# (i.e all fold id based separation is taken off).
sorted_srl_correct = sorted(srl_all_data.items(), key=lambda x: x[1][1][1], reverse=True)
srl_yield = []
gold_role_total = 0
gold_role_predicted = 0
total_role_predicted = 0
for x in sorted_srl_correct:
key, data = x
gold_role, srl_data = data
srl_role, srl_score = srl_data
gold_role_total += 1
if srl_role == gold_role:
gold_role_predicted += 1
total_role_predicted += 1
if gold_role_predicted != 0 and total_role_predicted != 0:
precision = gold_role_predicted/float(total_role_predicted)
else:
precision = 0
srl_yield.append((gold_role_predicted, gold_role_total, precision))
srl_df = pd.DataFrame(srl_yield)
srl_df.columns = ['yield', 'total_predicted', 'precision']
srl_df['recall'] = srl_df['yield']/max(srl_df.total_predicted.tolist())
srl_yield_df = srl_df.set_index(['recall'])
srl_yield_df = srl_yield_df['precision']
srl_plot_df = pd.DataFrame(srl_yield_df)
# call plot function
return srl_plot_df