/
knowledge_select.py
248 lines (236 loc) · 11.1 KB
/
knowledge_select.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
import torch
import numpy as np
import pickle
import nltk
from nltk import word_tokenize
from nltk.corpus import sentiwordnet as swn
emotion_dict = {'happiness': 0, 'neutral': 1, 'anger': 2, 'sadness': 3, 'fear': 4, 'surprise': 5, 'disgust': 6}
emotion_mapping = {0: 1, 1: 0, 2: -1, 3: -1, 4: -1, 5: -1, 6: -1}
senti_mapping = {-1: '-', 0: 'o', 1: '+'}
know_edge = {'o+': 1, 'o-': 2, '+': 3, '-': 4}
def text_score(text):
if text == '':
return []
if text == ' none':
return []
tag_seq = nltk.pos_tag([i for i in word_tokenize(str(text).lower())])
key = []
pos = []
n = ['NN', 'NNP', 'NNPS', 'NNS', 'UH']
v = ['VB', 'VBD', 'VBG', 'VBN', 'VBP', 'VBZ']
a = ['JJ', 'JJR', 'JJS']
r = ['RB', 'RBR', 'RBS', 'RP', 'WRB']
for i in range(len(tag_seq)):
key.append(tag_seq[i][0])
if tag_seq[i][1] in n:
pos.append('n')
elif tag_seq[i][1] in v:
pos.append('v')
elif tag_seq[i][1] in a:
pos.append('a')
elif tag_seq[i][1] in r:
pos.append('r')
else:
pos.append('')
pos_score = 0.
neg_score = 0.
obj_score = 0.
num_words = 0
for i in range(len(key)):
m = list(swn.senti_synsets(key[i], pos[i]))
if len(m) > 0:
num_words += 1
selected_synset = m[0]
pos_score = pos_score + selected_synset.pos_score()
neg_score = neg_score + selected_synset.neg_score()
obj_score = obj_score + selected_synset.obj_score()
if num_words != 0:
pos_score = pos_score / num_words
neg_score = neg_score / num_words
obj_score = obj_score / num_words
sen_score = pos_score - neg_score
sen_score = sen_score if abs(sen_score) > obj_score else 0.
if sen_score != 0.:
sen_score = 1 if sen_score > 0 else -1
return [text, sen_score]
else:
return []
def read_data(data_path, knowledge_path, processed_knowledge_path, window=10):
data = pickle.load(open(data_path, 'rb'), encoding='latin1')
emotion_label = data[3]
# evident_label = data[5]
knowledge = pickle.load(open(knowledge_path, 'rb'), encoding='latin1')
all_conversation_selected_knowledge = []
all_conversation_selected_edge_adj = []
for conv_id in range(len(emotion_label)):
conv_know = [' none</s></s>']
# conv_know = [' none ']
conversation_emotion = emotion_label[conv_id].numpy().tolist()
# conversation_evidence = evident_label[conv_id].numpy().tolist()
conversation_know = knowledge[conv_id]
conv_len = len(conversation_emotion)
edge_type_adj = np.zeros((conv_len, conv_len), dtype=np.int)
edge_index_adj = np.zeros((conv_len, conv_len), dtype=np.int)
edge_index = 1
for i in range(len(conversation_emotion)):
source_emotion = conversation_emotion[i]
source_sentiment = emotion_mapping[source_emotion]
source_senti_token = senti_mapping[source_sentiment]
source_knowledge = conversation_know[i]
f_item = min(conv_len, i+window+1)
source_xEffect = source_knowledge['xEffect'].split(' ==sep== ')
source_xEffect = [text_score(k) for k in source_xEffect]
source_oEffect = source_knowledge['oEffect'].split(' ==sep== ')
source_oEffect = [text_score(k) for k in source_oEffect]
source_xReact = source_knowledge['xReact'].split(' ==sep== ')
source_xReact = [text_score(k) for k in source_xReact]
source_oReact = source_knowledge['oReact'].split(' ==sep== ')
source_oReact = [text_score(k) for k in source_oReact]
source_same_speaker_know = {'o': ' none</s></s>', '+': ' none</s></s>', '-': ' none</s></s>'}
# source_same_speaker_know = {'o': ' none ', '+': ' none ', '-': ' none '}
neu_know = ''
pos_know = ''
neg_know = ''
for item in source_xEffect:
if len(item) == 0:
continue
kn, sent = item
if sent == 0:
neu_know = neu_know + kn + '</s></s>'
# neu_know = neu_know + kn
elif sent == 1:
pos_know = pos_know + kn + '</s></s>'
# pos_know = pos_know + kn
elif sent == -1:
neg_know = neg_know + kn + '</s></s>'
# neg_know = neg_know + kn
else:
raise NotImplementedError()
for item in source_xReact:
if len(item) == 0:
continue
kn, sent = item
if sent == 0:
neu_know = neu_know + kn + '</s></s>'
# neu_know = neu_know + kn
elif sent == 1:
pos_know = pos_know + kn + '</s></s>'
# pos_know = pos_know + kn
elif sent == -1:
neg_know = neg_know + kn + '</s></s>'
# neg_know = neg_know + kn
else:
raise NotImplementedError()
if neu_know != '':
source_same_speaker_know['o'] = neu_know
if pos_know != '':
source_same_speaker_know['+'] = pos_know
if neg_know != '':
source_same_speaker_know['-'] = neg_know
source_different_speaker_know = {'o': ' none</s></s>', '+': ' none</s></s>', '-': ' none</s></s>'}
# source_different_speaker_know = {'o': ' none ', '+': ' none ', '-': ' none '}
neu_know = ''
pos_know = ''
neg_know = ''
for item in source_oEffect:
if len(item) == 0:
continue
kn, sent = item
if sent == 0:
neu_know = neu_know + kn + '</s></s>'
# neu_know = neu_know + kn
elif sent == 1:
pos_know = pos_know + kn + '</s></s>'
# pos_know = pos_know + kn
elif sent == -1:
neg_know = neg_know + kn + '</s></s>'
# neg_know = neg_know + kn
else:
raise NotImplementedError()
for item in source_oReact:
if len(item) == 0:
continue
kn, sent = item
if sent == 0:
neu_know = neu_know + kn + '</s></s>'
# neu_know = neu_know + kn
elif sent == 1:
pos_know = pos_know + kn + '</s></s>'
# pos_know = pos_know + kn
elif sent == -1:
neg_know = neg_know + kn + '</s></s>'
# neg_know = neg_know + kn
else:
raise NotImplementedError()
if neu_know != '':
source_different_speaker_know['o'] = neu_know
if pos_know != '':
source_different_speaker_know['+'] = pos_know
if neg_know != '':
source_different_speaker_know['-'] = neg_know
same_speaker = True
edge_buffer = {}
for j in range(i, f_item):
target_emotion = conversation_emotion[j]
target_sentiment = emotion_mapping[target_emotion]
target_senti_token = senti_mapping[target_sentiment]
if source_senti_token == 'o':
interaction = source_senti_token + target_senti_token
else:
interaction = target_senti_token
if interaction in know_edge:
edge_type_adj[j, i] = know_edge[interaction]
if same_speaker:
interaction_speaker = interaction + 's'
if interaction_speaker in edge_buffer:
edge_index_adj[j, i] = edge_buffer[interaction_speaker]
else:
if source_senti_token == 'o':
klg = source_same_speaker_know['o'] + source_same_speaker_know[target_senti_token]
else:
klg = source_same_speaker_know[target_senti_token]
klg = source_same_speaker_know[target_senti_token]
if klg.strip(' ') != conv_know[0].strip(' '):
conv_know.append(klg)
edge_buffer[interaction_speaker] = edge_index
edge_index += 1
else:
edge_buffer[interaction_speaker] = 0
edge_index_adj[j, i] = edge_buffer[interaction_speaker]
same_speaker = False
else:
interaction_speaker = interaction + 'd'
if interaction_speaker in edge_buffer:
edge_index_adj[j, i] = edge_buffer[interaction_speaker]
else:
if source_senti_token == 'o':
klg = source_different_speaker_know['o'] + source_different_speaker_know[target_senti_token]
else:
klg = source_different_speaker_know[target_senti_token]
klg = source_different_speaker_know[target_senti_token]
if klg != conv_know[0]:
conv_know.append(klg)
edge_buffer[interaction_speaker] = edge_index
edge_index += 1
else:
edge_buffer[interaction_speaker] = 0
edge_index_adj[j, i] = edge_buffer[interaction_speaker]
same_speaker = True
else:
edge_type_adj[j, i] = 0
edge_index_adj[j, i] = 0
all_conversation_selected_knowledge.append(conv_know)
all_conversation_selected_edge_adj.append(edge_index_adj)
print(all_conversation_selected_knowledge[0])
print(all_conversation_selected_edge_adj[0])
pickle.dump([all_conversation_selected_knowledge, all_conversation_selected_edge_adj], open(processed_knowledge_path, 'wb'))
if __name__ == '__main__':
read_data('dd_data/dailydialog_train_processed.pkl',
'dd_data/dailydialog_train_know.pkl',
'dd_data/dailydialog_train_know_processed.pkl')
read_data('dd_data/dailydialog_dev_processed.pkl',
'dd_data/dailydialog_dev_know.pkl',
'dd_data/dailydialog_dev_know_processed.pkl')
read_data('dd_data/dailydialog_test_processed.pkl',
'dd_data/dailydialog_test_know.pkl',
'dd_data/dailydialog_test_know_processed.pkl')