/
extract_sentiment.py
98 lines (74 loc) · 4.58 KB
/
extract_sentiment.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
from read_write_file import *
from transformers import AutoModelForSequenceClassification
from transformers import TFAutoModelForSequenceClassification
from transformers import AutoTokenizer, AutoConfig
import numpy as np
from scipy.special import softmax
def extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = ''):
tokenizer = AutoTokenizer.from_pretrained(model_name)
#config = AutoConfig.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)
dataset = load_list_from_jsonl_file(input_file)
label_list = ['negative', 'neutral', 'positive']
for item in dataset:
source = item['source']
candidate = item['best_can']
target = item['target']
# source
encoded_input = tokenizer(source, return_tensors='pt', max_length = 512)
output = model(**encoded_input)
scores = output[0][0].detach().numpy()
scores = list(softmax(scores))
source_dict = {}
for label, value in zip(label_list, scores):
source_dict[label] = value
item['source_dict'] = source_dict
# candidate
encoded_input = tokenizer(candidate, return_tensors='pt', max_length = 512)
output = model(**encoded_input)
scores = output[0][0].detach().numpy()
scores = list(softmax(scores))
candidate_dict = {}
for label, value in zip(label_list, scores):
candidate_dict[label] = value
item['candidate_dict'] = candidate_dict
# target
encoded_input = tokenizer(target, return_tensors='pt', max_length = 512)
output = model(**encoded_input)
scores = output[0][0].detach().numpy()
scores = list(softmax(scores))
target_dict = {}
for label, value in zip(label_list, scores):
target_dict[label] = value
item['target_dict'] = target_dict
print('item: ', item)
print('-------------------------------')
# save
write_list_to_jsonl_file(input_file, dataset, file_access = 'w')
#....................
if __name__ == '__main__':
'''extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_test_para_256_diff_sim_best.json')'''
'''extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_validation_para_256_diff_sim_best.json')'''
'''extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_test_para_256_diff_rouge_best.json')'''
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_validation_para_256_diff_rouge_best.json')
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_test_para_256_diff_simrouge_best.json')
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_validation_para_256_diff_simrouge_best.json')
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_test_para_256_random_sim_best.json')
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_validation_para_256_random_sim_best.json')
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_test_para_256_random_rouge_best.json')
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_validation_para_256_random_rouge_best.json')
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_test_para_256_random_simrouge_best.json')
extract_dataset(model_name = 'cardiffnlp/twitter-xlm-roberta-base-sentiment', \
input_file = 'dataset/phrase2/generated_validation_para_256_random_simrouge_best.json')