-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathgraph_process.py
160 lines (142 loc) · 6.04 KB
/
graph_process.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
import pickle
import numpy as np
# data = pickle.load(open('bert_data/IEMOCAP/IEMOCAP_data.pkl', 'rb'), encoding='utf-8')
# train_data, test_data = data['train'], data['test']
#
# train_text, train_label = train_data[0], train_data[1]
# test_text, test_label = test_data[0], test_data[1]
#
# train_spk = train_data[2]
# test_spk = test_data[2]
def get_graph(spk_list, max_hip=1):
edge_index = []
edge_types = []
for conv in spk_list:
edge_idx = []
edge_type = []
for i in range(len(conv)):
xwant_hip = 0
owant_hip = 0
xintent_hip = 0
s_i = conv[i]
j = i - 1
pre_edge_idx = []
pre_edge_type = []
while j >= 0:
if conv[j] == s_i and xintent_hip < max_hip:
pre_edge_idx.append([i, j])
pre_edge_type.append('xIntent')
xintent_hip += 1
if xintent_hip == max_hip:
break
j -= 1
for k in range(len(pre_edge_idx) - 1, -1, -1):
edge_idx.append(pre_edge_idx[k])
edge_type.append(pre_edge_type[k])
edge_idx.append([i, i])
edge_type.append('xEffect')
j = i + 1
while j < len(conv):
if conv[j] == s_i and xwant_hip < max_hip:
edge_idx.append([i, j])
edge_type.append('xWant')
xwant_hip += 1
if conv[j] != s_i and owant_hip < max_hip:
edge_idx.append([i, j])
edge_type.append('oWant')
owant_hip += 1
if xwant_hip == max_hip and owant_hip == max_hip:
break
j += 1
edge_index.append(np.transpose(np.array(edge_idx)))
edge_types.append(edge_type)
graph = {'edge_index': edge_index, 'edge_type': edge_types}
return graph
# train_graph = get_graph(train_spk, max_hip=10)
# test_graph = get_graph(test_spk, max_hip=10)
# data = {'train': [train_text, train_label, train_spk, train_graph],
# 'test': [test_text, test_label, test_spk, test_graph]}
# pickle.dump(data, open('bert_data/IEMOCAP/IEMOCAP_graph_hip10_new.pkl', 'wb'))
def get_graph_multi_speaker(spk_list, max_hip=1):
edge_index = []
edge_types = []
for conv in spk_list:
edge_idx = []
edge_type = []
num_spk = np.unique(conv)
for i in range(len(conv)):
xwant_hip = 0
owant_hip = [0 for spk in range(np.max(num_spk) + 1)]
xintent_hip = 0
s_i = conv[i]
j = i - 1
pre_edge_idx = []
pre_edge_type = []
while j >= 0:
if conv[j] == s_i and xintent_hip < max_hip:
pre_edge_idx.append([i, j])
pre_edge_type.append('xIntent')
xintent_hip += 1
if xintent_hip == max_hip:
break
j -= 1
for k in range(len(pre_edge_idx) - 1, -1, -1):
edge_idx.append(pre_edge_idx[k])
edge_type.append(pre_edge_type[k])
edge_idx.append([i, i])
edge_type.append('xEffect')
j = i + 1
while j < len(conv):
if conv[j] == s_i and xwant_hip < max_hip:
edge_idx.append([i, j])
edge_type.append('xWant')
xwant_hip += 1
if conv[j] != s_i:
if owant_hip[conv[j]] < max_hip:
edge_idx.append([i, j])
edge_type.append('oWant')
owant_hip[conv[j]] += 1
j += 1
edge_index.append(np.transpose(np.array(edge_idx)))
edge_types.append(edge_type)
graph = {'edge_index': edge_index, 'edge_type': edge_types}
return graph
data = pickle.load(open('bert_data/MELD/MELD_data.pkl', 'rb'), encoding='utf-8')
train_data, dev_data, test_data = data
train_utt, train_spk, train_label = train_data
dev_utt, dev_spk, dev_label = dev_data
test_utt, test_spk, test_label = test_data
train_graph = get_graph_multi_speaker(train_spk, 7)
dev_graph = get_graph_multi_speaker(dev_spk, 7)
test_graph = get_graph_multi_speaker(test_spk, 7)
data = {'train': [train_utt, train_label, train_spk, train_graph],
'dev': [dev_utt, dev_label, dev_spk, dev_graph],
'test': [test_utt, test_label, test_spk, test_graph]}
pickle.dump(data, open('bert_data/MELD/MELD_graph_hip7_new.pkl', 'wb'))
# data = pickle.load(open('bert_data/EmoryNLP/EmoryNLP_feature.pkl', 'rb'), encoding='utf-8')
# train_data, test_data, dev_data = data
# train_utt, train_spk, train_label = train_data
# dev_utt, dev_spk, dev_label = dev_data
# test_utt, test_spk, test_label = test_data
# train_graph = get_graph_multi_speaker(train_spk, 7)
# dev_graph = get_graph_multi_speaker(dev_spk, 7)
# test_graph = get_graph_multi_speaker(test_spk, 7)
# data = {'train': [train_utt, train_label, train_spk, train_graph],
# 'dev': [dev_utt, dev_label, dev_spk, dev_graph],
# 'test': [test_utt, test_label, test_spk, test_graph]}
# pickle.dump(data, open('bert_data/EmoryNLP/EmoryNLP_graph_hip7_new.pkl', 'wb'))
# data = pickle.load(open('bert_data/DailyDialog/DailyDialog_feature.pkl', 'rb'), encoding='utf-8')
# train_data, dev_data, test_data = data
# train_utt, train_label, train_spk = train_data
# dev_utt, dev_label, dev_spk = dev_data
# test_utt, test_label, test_spk = test_data
#
# train_graph = get_graph(train_spk, 9)
# dev_graph = get_graph(dev_spk, 9)
# test_graph = get_graph(test_spk, 9)
#
# data = {'train': [train_utt, train_label, train_spk, train_graph],
# 'dev': [dev_utt, dev_label, dev_spk, dev_graph],
# 'test': [test_utt, test_label, test_spk, test_graph]}
#
# pickle.dump(data, open('bert_data/DailyDialog/DailyDialog_graph_hip9_new.pkl', 'wb'))