-2/dataHelper.py at main · dwopdm/-2 · GitHub

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
import datasets
import transformers
import json
import itertools
import pandas as pd
from datasets import Dataset, DatasetDict

def get_subdataset(subdataset_name, sep_token, label_offset=0):

	traintextlist=[]
	trainlabellist=[]
	testtextlist=[]
	testlabellist=[]

	if 'restaurant' in subdataset_name or 'laptop' in subdataset_name:
		if('restaurant' in subdataset_name):
			train_file_path="./SemEval14-res/train.json"
			test_file_path="./SemEval14-res/test.json"

		elif('laptop' in subdataset_name):
			train_file_path="./SemEval14-laptop/train.json"
			test_file_path="./SemEval14-laptop/test.json"

		with open(train_file_path, 'r') as train_file:
			with open(test_file_path, 'r') as test_file:
				train=json.load(train_file)
				test=json.load(test_file)

				for value in (list)(train.values()):
					if value['polarity']=='positive':
						trainlabellist.append(0+label_offset)
					elif value['polarity']=='neutral':
						trainlabellist.append(1+label_offset)
					else:
						trainlabellist.append(2+label_offset)
					traintextlist.append(value['term']+' '+sep_token+value['sentence'])

				for value in (list)(test.values()):
					if value['polarity']=='positive':
						testlabellist.append(0+label_offset)
					elif value['polarity']=='neutral':
						testlabellist.append(1+label_offset)
					else:
						testlabellist.append(2+label_offset)
					testtextlist.append(value['term']+' '+sep_token+value['sentence'])

		train_dataset=Dataset.from_dict({'text': traintextlist, 'labels': trainlabellist})
		test_dataset=Dataset.from_dict({'text': testtextlist, 'labels': testlabellist})
		subdataset=DatasetDict(
			{
				'train': train_dataset,
				'test': test_dataset
			}
		)

	elif 'acl' in subdataset_name:
		train_file_path="./acl-arc/train.jsonl"
		test_file_path="./acl-arc/test.jsonl"
		with open(train_file_path, 'r') as train_file:
			with open(test_file_path, 'r') as test_file:
				for train_line in train_file:
					traintextlist.append(json.loads(train_line)['text'])
					trainlabellist.append(json.loads(train_line)['intent']+label_offset)
				for test_line in test_file:
					testtextlist.append(json.loads(test_line)['text'])
					testlabellist.append(json.loads(test_line)['intent']+label_offset)

		train_dataset=Dataset.from_dict({'text': traintextlist, 'labels': trainlabellist})
		test_dataset=Dataset.from_dict({'text': testtextlist, 'labels': testlabellist})
		subdataset=DatasetDict(
			{
				'train': train_dataset,
				'test': test_dataset
			}
		)

	elif 'agnews' in subdataset_name:
		Dict={}
		textlist=[]
		labellist=[]
		with open("./agnews/test.jsonl", 'r') as file:
			for line in file:
				textlist.append(json.loads(line)['text'])
				labellist.append(json.loads(line)['label']+label_offset)

		subdataset=Dataset.from_dict({'text': textlist, 'labels': labellist})
		subdataset=subdataset.train_test_split(test_size=0.1, seed=2022, shuffle=True)

	if 'fs' in subdataset_name:
		seed = 2022
		num_labels = max(subdataset['train']['labels']) - label_offset

		if num_labels < 4:
			subdataset['train'] = subdataset['train'].shuffle(seed=seed)
			subdataset['train'] = subdataset['train'].select(range(32))
		else:
			subdataset['train'] = subdataset['train'].shuffle(seed=seed)
			_idx = [[] for i in range(num_labels+1)]
			for idx, label in enumerate(subdataset['train']['labels']):
				if len(_idx[label]) < 8:
					_idx[label].append(idx)
			idx_lst = [i for item in _idx for i in item]
			subdataset['train'] = subdataset['train'].select(idx_lst).shuffle(seed=seed)

	return subdataset

def get_dataset(dataset_name, sep_token):
	'''
	dataset_name: str, the name of the dataset
	sep_token: str, the sep_token used by tokenizer(e.g. '<sep>')
	'''
	if type(dataset_name)!=list:
		print("1")
		return get_subdataset(dataset_name, sep_token)
	else:
		label_offset=0
		train_dataset_list=[]
		test_dataset_list=[]
		for dataset in dataset_name:
			subdataset=get_subdataset(dataset, sep_token, label_offset)
			label_offset=max(subdataset['train']['labels'])+1
			train_dataset_list.append(subdataset['train'])
			test_dataset_list.append(subdataset['test'])
		return DatasetDict({'train': datasets.concatenate_datasets(train_dataset_list),
					  		'test': datasets.concatenate_datasets(test_dataset_list)})

def main():
	test1=get_dataset(["agnews_fs", "restaurant_fs"], '<SEP>')
	train=test1['train']
	text=train['text']
	print(text[0])

if __name__ == '__main__':
	main()