/
main.py
215 lines (157 loc) · 6.31 KB
/
main.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
from hazm import word_tokenize, Lemmatizer
from numpy.random import choice
import string
import math
def load_corpus(data_path):
with open(data_path, "r", encoding="utf-8") as corpus:
corpus = corpus.read()
return corpus
def tokenize(corpus,lemma=True, punctuation=True, space_to_space=False):
if(not punctuation):
# table = str.maketrans({key: None for key in string.punctuation})
# corpus = corpus.translate(table)
corpus = corpus.replace(',', ' ')
corpus=corpus.replace("\u220c","")
corpus = corpus.replace('(', ' ')
corpus = corpus.replace(')', ' ')
corpus = corpus.replace('.', ' ')
corpus=corpus.replace("،"," ")
corpus=corpus.replace("«"," ")
corpus=corpus.replace("»"," ")
if(space_to_space):
tokenized = corpus.split(' ')
else:
tokenized = word_tokenize(corpus)
if(lemma):
lemmatizer=Lemmatizer()
for i in range(len(tokenized)):
tokenized[i] = lemmatizer.lemmatize(tokenized[i]).split('#')[0]
return tokenized
def generate_n_gram(tokenized, n):
ngrams_list = []
ngrams = []
for num in range(len(tokenized)):
ngram=(tokenized[num:num + n])
ngrams_list.append(ngram)
for i in range(len(ngrams_list)):
flag = False
for j in range(len(ngrams)):
if ngrams[j][0]==ngrams_list[i]:
flag = True
if not flag:
ngrams.append((ngrams_list[i],ngrams_list.count(ngrams_list[i])))
return ngrams
def generate_sentence(n, start_words, ngrams, ngrams_minus_1):
sentence = start_words
multiplied_probs =1
sum_log_probs = 0
i=0
while not sentence[-1]=='</S>':
options = []
probs = []
for ngram_1 in ngrams_minus_1:
if ngram_1[0]== sentence[i:i+n-1]:
count_ngram_minus_1 = ngram_1[1]
for ngram in ngrams:
if ngram[0][0:n-1]== sentence[i:i+n-1]:
options.append(ngram[0][n-1])
probs.append(ngram[1]/count_ngram_minus_1)
winner = choice(options, 1, probs)
sentence.append(winner[0])
multiplied_probs = multiplied_probs * probs[options.index(winner)]
sum_log_probs = sum_log_probs + math.log(probs[options.index(winner)],2)
i=i+1
#perplexity= math.pow((1/multiplied_probs),(1/len(sentence)))
perplexity_by_log= 2** (-1.0 * sum_log_probs / len(sentence))
# print(perplexity, perplexity_by_log)
return sentence, perplexity_by_log
def generate_all_sentences(tokenized):
sentences = []
result = ''
for n in range(2,6):
ngrams = generate_n_gram(tokenized, n)
ngrams_minus_1 = generate_n_gram(tokenized, n-1)
for j in range(10):
if(n==2):
sentence, perplexity_by_log = generate_sentence(n=n,start_words=['<S>'] ,ngrams=ngrams, ngrams_minus_1=ngrams_minus_1)
sentences.append(sentence)
if(not n==2):
sentence, perplexity_by_log = generate_sentence(n=n,start_words=sentences[(n-3)*10+j][0:n-1] ,ngrams=ngrams, ngrams_minus_1=ngrams_minus_1)
sentences.append(sentence)
# print('ngram: ', n, 'sentence: ', sentence, 'perplexity: ', perplexity, 'perplexity_by_log: ',perplexity_by_log)
result = result+ '\nn: '+ str(n)
result = result+'\t'+' '.join(sentence)
result = result+ '\nperplexity: ' + str(perplexity_by_log)
return result
def compute_test_perplexity(n, test_tokenized, tokenized):
ngrams = generate_n_gram(tokenized, n)
ngrams_minus_1 = generate_n_gram(tokenized, n-1)
sum_log_probs = 0
for i in range(len(test_tokenized)):
found = False
for ngram in ngrams:
if ngram[0] == test_tokenized[i-n:i]:
count_ngram = ngram[1]
found = True
if(not found):
prob = 1
else:
for ngram_1 in ngrams_minus_1:
if ngram_1[0] == test_tokenized[i-n:i-1]:
prob = count_ngram / ngram_1[1]
sum_log_probs = sum_log_probs + math.log(prob,2)
perplexity_by_log= 2** (-1.0 * sum_log_probs / len(test_tokenized))
print (perplexity_by_log)
return
def generate_unigram_sentences(tokenized):
unigrams = []
result = ''
vocab = set(tokenized)
for word in vocab:
unigrams.append((word,tokenized.count(word)/len(vocab)))
for j in range(100):
sentence = []
i=0
sum_log_probs = 0
while ( (len(sentence)==0 or not sentence[-1]=='</S>') and len(sentence)<15) :
options=[i[0] for i in unigrams]
probs=[i[1] for i in unigrams]
winner = choice(options, 1, probs)
sentence.append(winner[0])
sum_log_probs = sum_log_probs + math.log(probs[options.index(winner)],2)
i=i+1
perplexity_by_log= 2** (-1.0 * sum_log_probs / len(sentence))
result = result+ '\nn: 1'
result = result+'\t'+' '.join(sentence)
result = result+ '\nperplexity: ' + str(perplexity_by_log)
return result
def compute_unigram_test_perplexity(test_tokenized, tokenized):
unigrams = []
vocab = set(tokenized)
for word in vocab:
unigrams.append((word,tokenized.count(word)/len(vocab)))
sum_log_probs = 0
for i in range(len(test_tokenized)):
for unigram in unigrams:
if(unigram[0]==test_tokenized[i]):
prob = unigram[1]
sum_log_probs = sum_log_probs + math.log(prob,2)
perplexity_by_log= 2** (-1.0 * sum_log_probs / len(test_tokenized))
print (perplexity_by_log)
return
# load and token
corpus = load_corpus("train.txt")
tokenized = tokenize(corpus,lemma=False, punctuation=False)
# generate sentences n-gram
# result = generate_all_sentences(tokenized)
# result = generate_unigram_sentences(tokenized)
# f= open("ngram_sentenses.txt","w+")
# f.write(result)
# f.close()
## Test
corpus = load_corpus("train3.txt")
tokenized = tokenize(corpus,lemma=True, punctuation=True)
corpus = load_corpus("test.txt")
test_tokenized = tokenize(corpus,lemma=True, punctuation=True)
# compute_test_perplexity(n=4, test_tokenized=test_tokenized, tokenized=tokenized)
compute_unigram_test_perplexity(test_tokenized, tokenized)