-
Notifications
You must be signed in to change notification settings - Fork 2
/
reformat3.py
231 lines (213 loc) · 15.1 KB
/
reformat3.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
import string
import gzip
import os
import sys
import time
import numpy
import theano.sandbox.neighbours as TSN
import theano
import theano.tensor as T
from theano.tensor.signal import downsample
from theano.tensor.nnet import conv
from logistic_sgd import LogisticRegression
from mlp import HiddenLayer
from cis.deep.utils.theano import debug_print
numcontext = 10
class trainex:
def triple(self):
#print (self.word,self.left,self.rght)
return (self.word,self.left,self.rght)
#print self.word, self.left, self.rght
def __init__(self,mywords,i,sent,i2w):
self.word = i2w[mywords[i]]
self.left = []
j = i
while len(self.left)<numcontext:
j -= 1
if j<0:
self.left.append('padd')
elif mywords[j]==mywords[i]:
continue
else:
self.left.append(sent[j])
self.left.reverse()
self.rght = []
j = i
while len(self.rght)<numcontext:
j += 1
if j>=len(sent):
self.rght.append('padd')
elif mywords[j]==mywords[i]:
continue
else:
self.rght.append(sent[j])
class wikiline:
def sentence(self):
return self.sent
def triples(self):
for myex in self.alltrainex:
myex.triple()
def __init__(self,myline):
myparts = string.split(myline,'5')
self.sent = []
self.alltrainex = []
mywords = []
i2w = {}
for i,myword in enumerate(myparts):
if myword=='': continue
subparts = string.split(myword,'6')
#print subparts
assert len(subparts)==2
i2w[i] = string.strip(subparts[0])
subsubparts = string.split(subparts[1])
for subsubpart in subsubparts:
if subsubpart=='': continue
self.sent.append(subsubpart)
mywords.append(i)
done = set()
for i in range(len(self.sent)):
if mywords[i] in done: continue
done.add(mywords[i])
self.alltrainex.append(trainex(mywords,i,self.sent,i2w))
def yinwikireformat3(maxlength, window_size, max_size):
#defined by wenpeng
trigram2id={}
id2trigram={}
target_word2id={}
target_id2word={}
wordid2count={}
trigram_count=0
word_count=0
all_word_numbers=0
data=[]
context_matrix=[]
target_matrix=[]
Lengths=[]
Lengths_target=[]
leftPad=[]
rightPad=[]
filename = '/mounts/Users/cisintern/hs/l/schuetze2014/yin/cnnlm/wiki,ebert,uniq,ngram.txt'
myfile = open(filename,'r')
count=0
for myline in myfile:
#print myline
myline = string.strip(myline)
if myline=='': continue
myobj = wikiline(myline)
'''
print
print myobj.sentence()
#myobj.triples()
'''
input_length=len(myobj.sentence())
#print 'input_length: '+str(input_length)
if input_length>maxlength or input_length<20:
continue
else: # a valid sentence
if count==max_size:
break
count+=1
sent=[]
contexts=[]
targets=[]
targetWords_per_sentence=0
Lengths.append(input_length)
left=(maxlength-input_length)/2
right=maxlength-left-input_length
leftPad.append(left)
rightPad.append(right)
sent+=[0]*left
for trigram in myobj.sentence():
#print trigram
id=trigram2id.get(trigram, -1) # possibly the embeddings are for words with lowercase
if id == -1:
#embeddings.append(numpy.random.uniform(-1,1,embedding_size)) # generate a random embedding for an unknown word
#embeddings_target.append(numpy.random.uniform(-1,1,embedding_size))
trigram2id[trigram]=trigram_count+1 # starts from 1
id2trigram[trigram_count+1]=trigram #1 means new words
sent.append(trigram_count+1)
trigram_count+=1
else:
sent.append(id)
sent+=[0]*right
#left context
for myex in myobj.alltrainex:
target_word, left_contexts, right_contexts=myex.triple()
#print target_word, left_contexts, right_contexts
targetWords_per_sentence+=1
#store context for target word
for context in left_contexts:
#print context
id=trigram2id.get(context,0) #padd is set to index 0
contexts.append(id)
if id==0:
id2trigram[id]=context
for context in right_contexts:
id=trigram2id.get(context,0) #padd is set to index 0
contexts.append(id)
if id==0:
id2trigram[id]=context
#store target word
all_word_numbers+=1
id=target_word2id.get(target_word,-1)
if id==-1: #a new word
target_word2id[target_word]=word_count # word index starts from 0
target_id2word[word_count]=target_word
id=word_count
wordid2count[id]=1 #the first time to appear
word_count+=1
else:
wordid2count[id]=wordid2count[id]+1
targets.append(id)
data.append(sent)
repeated_targets=targets*60
target_matrix.append(repeated_targets[:60]) #only consider maxmum 60 target words for training
repeated_context=contexts*60
context_matrix.append(repeated_context[:(60*window_size*2)]) # consider 60*(context_size)
Lengths_target.append(targetWords_per_sentence)
'''
print 'data'
print data
print 'target_matrix'
print target_matrix
print 'context_matrix'
print context_matrix
print 'Lengths_target'
print Lengths_target
'''
print 'Wiki corpus loaded over. Totally '+str(word_count)+' distinct target words'
unigram=[]
for index in xrange(word_count):
unigram.append(wordid2count[index]*1.0/all_word_numbers)
#print unigram
def shared_dataset(data_y, borrow=True):
shared_y = theano.shared(numpy.asarray(data_y,
dtype=theano.config.floatX), # @UndefinedVariable
borrow=borrow)
return T.cast(shared_y, 'int32')
#return shared_y
train_set_Lengths=debug_print(shared_dataset(numpy.array(Lengths)),'train_set_length' )
#valid_set_Lengths = shared_dataset(devLengths)
#uni_gram=shared_dataset(unigram)
train_left_pad=debug_print(shared_dataset(numpy.array(leftPad)),'leftPad')
train_right_pad=debug_print(shared_dataset(numpy.array(rightPad)), 'rightPad')
#dev_left_pad=shared_dataset(devLeftPad)
#dev_right_pad=shared_dataset(devRightPad)
'''
print 'length_target:'
print Lengths_target
print 'context_matrix:'
print context_matrix
print 'target_matrix:'
print target_matrix
'''
if trigram_count+1!=len(id2trigram):
print 'trigram_count: '+str(trigram_count)
print 'id2trigram:'+str(len(id2trigram))
exit(0)
if len(data)!=len(leftPad) or len(data)!=len(Lengths) or len(data)!=len(Lengths_target) or len(data)!=len(context_matrix) or len(data)!=len(target_matrix):
print 'Load data error: sentence amount not equal to padding.'
exit(0)
rval = [(numpy.array(data),train_set_Lengths, train_left_pad, train_right_pad), (numpy.array(data),train_set_Lengths, train_left_pad, train_right_pad)]
return rval, numpy.array(unigram), numpy.array(Lengths), numpy.array(Lengths_target),trigram_count, numpy.array(context_matrix), numpy.array(target_matrix), target_id2word, id2trigram
#yinwikireformat3(250, 10)