-
Notifications
You must be signed in to change notification settings - Fork 0
/
shannon_info.py
208 lines (168 loc) · 5.83 KB
/
shannon_info.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
import numpy as np
phrasesNP = np.loadtxt('dataset/phrases.txt', dtype='str' , delimiter="\n")
phrases = list(phrasesNP)
from nltk.tokenize import RegexpTokenizer
tokenizer = RegexpTokenizer(r'\w+')
alls = " ".join(phrases)
wordlist = tokenizer.tokenize(alls)
uwordlist = set(wordlist)
dict_words = dict.fromkeys(uwordlist,0)
for i in wordlist:
dict_words[i] += 1
prob = {}
for i in dict_words:
prob[i] = dict_words[i] / sum(dict_words.values())
# “Shannon information,” “self-information,” or simply the “information,”
# information(x) = -log( p(x) )
import math
theinfo = -math.log(prob['the'],2)
info = {}
for i in dict_words:
info[i] = -math.log(prob[i],2)
''' *************************************************** '''
from better_profanity import profanity
profanity.load_censor_words()
i=-1
i+=1
phrase = hate[i]
p2=profanity.censor(phrase)
h = tokenizer.tokenize(phrase)
g = tokenizer.tokenize(p2)
l = list(set(h)-set(g))
s_info = [info[x.lower()] for x in l]
sum(s_info)
phraseO = phrase
phrase = p2
#phrase = phrase.replace('**** ','')
pra = get_response(phrase, num_return_sequences=int(x), num_beams=int(y), groups=int(z), diversityP=t)
results = modelD.predict(pra)
tox = np.array(results['toxicity'])
ix = np.where(tox < 0.5 )[0]
len(ix)
praGood = list(np.array(pra)[ix])
sim = similarity(phraseO, praGood)[0]
ixSim = np.where(sim>0.57)[0]
len(ixSim)
print(phrase)
print(phraseO)
print_list(np.array(praGood)[ixSim])
phrase.count('****') / len(phrase.split())
sum(s_info)
''' *************************calc info of censored words**************************** '''
#end=0
#while True:
# start = p2.find('****', end)
# end = phrase.find(' ',start)
# print(start, end)
# word = ''
# if end == -1:
# word = phrase[start:]
# break;
# else:
# word = phrase[start:end]
# print(info[word.lower()])
h = tokenizer.tokenize(phrase)
g = tokenizer.tokenize(p2)
l = list(set(h)-set(g))
i = [info[x.lower()] for x in l]
''' **********************analysis of info per POS tag******************************* '''
from nltk.tag.stanford import StanfordPOSTagger
st = StanfordPOSTagger('english-bidirectional-distsim.tagger')
#from nltk.tokenize import RegexpTokenizer
#tokenizer = RegexpTokenizer(r'\w+')
pos_info = {}
pos_count = {}
for h in hate:
splt = tokenizer.tokenize(h)
tagged = st.tag(splt)
## censored:
p2 = profanity.censor(h)
p2token = tokenizer.tokenize(p2)
intersect = list(set(splt)-set(p2token))
for e in tagged:
tag = e[1]
word = e[0]
if word not in intersect:
continue;
w_info = 0.0
try:
w_info = info[word.lower()]
except KeyError:
continue;
if tag == 'FW':
print(h, word, w_info)
if tag in pos_info:
pos_info[tag] += w_info
pos_count[tag] += 1
else:
pos_info[tag] = w_info
pos_count[tag] = 1
i_per_pos = {}
for tg in pos_info:
f = pos_info[tg]/pos_count[tg]
#print(tg, f)
i_per_pos[tg] = f
#all_sorted = {k: v for k, v in sorted(i_per_pos.items(), key=lambda item: item[1])}
relevant_tags = {k:i_per_pos[k] for k,v in pos_count.items() if v>30}
relevant_sorted = {k: v for k, v in sorted(relevant_tags.items(), key=lambda item: item[1])}
# {'DT': 6.23161312176529, 'PRP': 6.663108188877318, 'IN': 8.024913727923057, 'RB': 9.395283695147434, 'VBP': 9.920054652224884, 'VB': 10.13504759704436, 'JJ': 11.282893558661753, 'NN': 11.979298992357021, 'NNS': 12.021336296417894}
# determiner , personal pronoun , preposition , adverb , verb , adjective , noun
''' censored '''
#>>> pos_info
#{'NNS': 156.75022004436792, 'VBG': 20.88505777349939, 'VBD': 9.810621449838727, 'NN': 524.6288146731758, 'VB': 110.04383509127192, 'JJ': 201.9265734869597, 'VBP': 49.267278475478435, 'VBN': 59.51416286625181, 'FW': 65.64212194977752, 'RB': 6.3613928089523455, 'VBZ': 12.236886204540825, 'NNP': 32.25944750179015, 'WRB': 8.793942708692096, 'TO': 5.500484273222535}
#>>> pos_count
#{'NNS': 12, 'VBG': 2, 'VBD': 1, 'NN': 48, 'VB': 12, 'JJ': 20, 'VBP': 5, 'VBN': 6, 'FW': 7, 'RB': 1, 'VBZ': 1, 'NNP': 3, 'WRB': 1, 'TO': 1}
# verb, adjective, noun
#>>> i_per_pos
#{'NNS': 13.06251833703066, 'VBG': 10.442528886749695, 'VBD': 9.810621449838727, 'NN': 10.92976697235783, 'VB': 9.170319590939327, 'JJ': 10.096328674347985, 'VBP': 9.853455695095686, 'VBN': 9.9190271443753, 'FW': 9.37744599282536, 'RB': 6.3613928089523455, 'VBZ': 12.236886204540825, 'NNP': 10.753149167263382, 'WRB': 8.793942708692096, 'TO': 5.500484273222535}
''' c. corrected '''
#>>> join_count
#{'NN': 66, 'VB': 36, 'JJ': 15, 'RB': 9, 'MD': 1, 'WR': 1, 'TO': 1}
#>>> i_per_pos_join
#{'NN': 10.980905868262893, 'VB': 9.523328078596386, 'JJ': 9.843543969991522, 'RB': 9.151486294218174, 'MD': 12.49992061037462, 'WR': 8.793942708692096, 'TO': 5.500484273222535}
# adjective, adverb -> descriptive non essential for sentence generation
# FW foreign word -> check what is classified as foreign and if it can be deleted
# verb, noun - to important/essential
''' **********************analysis of info per POS tag - corrected (decontracted + nltk.pos_tag)******************************* '''
pos_info = {}
pos_count = {}
for h in hate:
dph = decontracted(h.replace("’","'"))
splt = tokenizer.tokenize(dph)
tagged = pos_tag(splt)
## censored:
p2 = profanity.censor(h)
p2token = tokenizer.tokenize(p2)
intersect = list(set(splt)-set(p2token))
for e in tagged:
tag = e[1]
word = e[0]
if word not in intersect:
continue;
w_info = 0.0
try:
w_info = info[word.lower()]
except KeyError:
continue;
if tag == 'FW':
print(h, word, w_info)
if tag in pos_info:
pos_info[tag] += w_info
pos_count[tag] += 1
else:
pos_info[tag] = w_info
pos_count[tag] = 1
join_i = {}
join_c = {}
for tag in pos_info:
new_tag = tag[:2]
if new_tag in join_i:
join_i[new_tag] += pos_info[tag]
join_c[new_tag] += pos_count[tag]
else:
join_i[new_tag] = pos_info[tag]
join_c[new_tag] = pos_count[tag]
i_per_pos_join = {}
for tg in join_i:
f = join_i[tg]/join_c[tg]
i_per_pos_join[tg]=f