Datasets:
update data loader
Browse files- README.md +104 -3
- kobest_v1.py +32 -44
README.md
CHANGED
@@ -65,15 +65,116 @@ Boolean Question Answering, Choice of Plausible Alternatives, Words-in-Context,
|
|
65 |
|
66 |
### Data Instances
|
67 |
|
68 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
### Data Fields
|
71 |
|
72 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
73 |
|
74 |
### Data Splits
|
75 |
|
76 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
77 |
|
78 |
## Dataset Creation
|
79 |
|
|
|
65 |
|
66 |
### Data Instances
|
67 |
|
68 |
+
#### KB-BoolQ
|
69 |
+
An example of a data point looks as follows.
|
70 |
+
```
|
71 |
+
{'paragraph': '두아 리파(Dua Lipa, 1995년 8월 22일 ~ )는 잉글랜드의 싱어송라이터, 모델이다. BBC 사운드 오브 2016 명단에 노미닛되었다. 싱글 "Be the One"가 영국 싱글 차트 9위까지 오르는 등 성과를 보여주었다.',
|
72 |
+
'question': '두아 리파는 영국인인가?',
|
73 |
+
'label': 1}
|
74 |
+
```
|
75 |
+
|
76 |
+
#### KB-COPA
|
77 |
+
An example of a data point looks as follows.
|
78 |
+
```
|
79 |
+
{'premise': '물을 오래 끓였다.',
|
80 |
+
'question': '결과',
|
81 |
+
'alternative_1': '물의 양이 늘어났다.',
|
82 |
+
'alternative_2': '물의 양이 줄어들었다.',
|
83 |
+
'label': 1}
|
84 |
+
```
|
85 |
+
|
86 |
+
#### KB-WiC
|
87 |
+
An example of a data point looks as follows.
|
88 |
+
```
|
89 |
+
{'word': '양분',
|
90 |
+
'context_1': '토양에 [양분]이 풍부하여 나무가 잘 자란다. ',
|
91 |
+
'context_2': '태아는 모체로부터 [양분]과 산소를 공급받게 된다.',
|
92 |
+
'label': 1}
|
93 |
+
```
|
94 |
+
|
95 |
+
#### KB-HellaSwag
|
96 |
+
An example of a data point looks as follows.
|
97 |
+
```
|
98 |
+
{'context': '모자를 쓴 투수가 타자에게 온 힘을 다해 공을 던진다. 공이 타자에게 빠른 속도로 다가온다. 타자가 공을 배트로 친다. 배트에서 깡 소리가 난다. 공이 하늘 위로 날아간다.',
|
99 |
+
'ending_1': '외야수가 떨어지는 공을 글러브로 잡는다.',
|
100 |
+
'ending_2': '외야수가 공이 떨어질 위치에 자리를 잡는다.',
|
101 |
+
'ending_3': '심판이 아웃을 외친다.',
|
102 |
+
'ending_4': '외야수가 공을 따라 뛰기 시작한다.',
|
103 |
+
'label': 3}
|
104 |
+
```
|
105 |
+
|
106 |
+
#### KB-SentiNeg
|
107 |
+
An example of a data point looks as follows.
|
108 |
+
```
|
109 |
+
{'sentence': '택배사 정말 마음에 듬',
|
110 |
+
'label': 1}
|
111 |
+
```
|
112 |
|
113 |
### Data Fields
|
114 |
|
115 |
+
### KB-BoolQ
|
116 |
+
+ `paragraph`: a `string` feature
|
117 |
+
+ `question`: a `string` feature
|
118 |
+
+ `label`: a classification label, with possible values `False`(0) and `True`(1)
|
119 |
+
|
120 |
+
|
121 |
+
### KB-COPA
|
122 |
+
+ `premise`: a `string` feature
|
123 |
+
+ `question`: a `string` feature
|
124 |
+
+ `alternative_1`: a `string` feature
|
125 |
+
+ `alternative_2`: a `string` feature
|
126 |
+
+ `label`: an answer candidate label, with possible values `alternative_1`(0) and `alternative_2`(1)
|
127 |
+
|
128 |
+
|
129 |
+
### KB-WiC
|
130 |
+
+ `target_word`: a `string` feature
|
131 |
+
+ `context_1`: a `string` feature
|
132 |
+
+ `context_2`: a `string` feature
|
133 |
+
+ `label`: a classification label, with possible values `False`(0) and `True`(1)
|
134 |
+
|
135 |
+
### KB-HellaSwag
|
136 |
+
+ `target_word`: a `string` feature
|
137 |
+
+ `context_1`: a `string` feature
|
138 |
+
+ `context_2`: a `string` feature
|
139 |
+
+ `label`: a classification label, with possible values `False`(0) and `True`(1)
|
140 |
+
|
141 |
+
### KB-SentiNeg
|
142 |
+
+ `sentence`: a `string` feature
|
143 |
+
+ `label`: a classification label, with possible values `Negative`(0) and `Positive`(1)
|
144 |
+
|
145 |
|
146 |
### Data Splits
|
147 |
|
148 |
+
#### KB-BoolQ
|
149 |
+
|
150 |
+
+ train: 3,665
|
151 |
+
+ dev: 700
|
152 |
+
+ test: 1,404
|
153 |
+
|
154 |
+
#### KB-COPA
|
155 |
+
|
156 |
+
+ train: 3,076
|
157 |
+
+ dev: 1,000
|
158 |
+
+ test: 1,000
|
159 |
+
|
160 |
+
#### KB-WiC
|
161 |
+
|
162 |
+
+ train: 3,318
|
163 |
+
+ dev: 1,260
|
164 |
+
+ test: 1,260
|
165 |
+
|
166 |
+
#### KB-HellaSwag
|
167 |
+
|
168 |
+
+ train: 3,665
|
169 |
+
+ dev: 700
|
170 |
+
+ test: 1,404
|
171 |
+
|
172 |
+
#### KB-SentiNeg
|
173 |
+
|
174 |
+
+ train: 3,649
|
175 |
+
+ dev: 400
|
176 |
+
+ test: 397
|
177 |
+
+ test_originated: 397 (Corresponding training data where the test set is originated from.)
|
178 |
|
179 |
## Dataset Creation
|
180 |
|
kobest_v1.py
CHANGED
@@ -34,6 +34,7 @@ _DATA_URLS = {
|
|
34 |
"train": _URL + "/v1.0/SentiNeg/train.tsv",
|
35 |
"dev": _URL + "/v1.0/SentiNeg/dev.tsv",
|
36 |
"test": _URL + "/v1.0/SentiNeg/test.tsv",
|
|
|
37 |
},
|
38 |
"hellaswag": {
|
39 |
"train": _URL + "/v1.0/HellaSwag/train.tsv",
|
@@ -69,26 +70,6 @@ class KoBESTConfig(datasets.BuilderConfig):
|
|
69 |
self.url = url
|
70 |
|
71 |
|
72 |
-
# class KoBESTConfig(datasets.BuilderConfig):
|
73 |
-
# """BuilderConfig for KoTEST."""
|
74 |
-
#
|
75 |
-
# def __init__(
|
76 |
-
# self,
|
77 |
-
# features,
|
78 |
-
# data_url,
|
79 |
-
# file_map,
|
80 |
-
# url,
|
81 |
-
# **kwargs,
|
82 |
-
# ):
|
83 |
-
# """BuilderConfig for KoTEST."""
|
84 |
-
#
|
85 |
-
# super(KoBESTConfig, self).__init__(version=datasets.Version("1.0.0", ""), **kwargs)
|
86 |
-
# self.features = features
|
87 |
-
# self.data_url = data_url
|
88 |
-
# self.file_map = file_map
|
89 |
-
# self.url = url
|
90 |
-
|
91 |
-
|
92 |
class KoBEST(datasets.GeneratorBasedBuilder):
|
93 |
BUILDER_CONFIGS = [
|
94 |
KoBESTConfig(name=name, description=_DESCRIPTION, data_url=_DATA_URLS[name], citation=_CITATAION, url=_URL)
|
@@ -144,6 +125,16 @@ class KoBEST(datasets.GeneratorBasedBuilder):
|
|
144 |
dev = dl_manager.download_and_extract(self.config.data_url["dev"])
|
145 |
test = dl_manager.download_and_extract(self.config.data_url["test"])
|
146 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
147 |
return [
|
148 |
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": train, "split": "train"}),
|
149 |
datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": dev, "split": "dev"}),
|
@@ -195,16 +186,6 @@ class KoBEST(datasets.GeneratorBasedBuilder):
|
|
195 |
df = df.dropna()
|
196 |
df = df[['context', 'choice1', 'choice2', 'choice3', 'choice4', 'label']]
|
197 |
|
198 |
-
# for id_, row in df.iterrows():
|
199 |
-
# yield id_, {
|
200 |
-
# "context": str(row["context"]),
|
201 |
-
# "ending_1": str(row["choice1"]),
|
202 |
-
# "ending_2": str(int(row["choice2"])),
|
203 |
-
# "ending_3": str(int(row["choice3"])),
|
204 |
-
# "ending_4": str(int(row["choice4"])),
|
205 |
-
# "label": str(row["label"]),
|
206 |
-
# }
|
207 |
-
#
|
208 |
df = df.rename(columns={
|
209 |
'context': 'context',
|
210 |
'choice1': 'ending_1',
|
@@ -217,18 +198,21 @@ class KoBEST(datasets.GeneratorBasedBuilder):
|
|
217 |
elif self.config.name == "sentineg":
|
218 |
df = pd.read_csv(filepath, sep="\t")
|
219 |
df = df.dropna()
|
220 |
-
df = df[['Text', 'Label']]
|
221 |
|
222 |
-
|
223 |
-
|
224 |
-
# "sentence": str(row["Text"]),
|
225 |
-
# "label": str(int(row["Label"])),
|
226 |
-
# }
|
227 |
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
232 |
|
233 |
else:
|
234 |
raise NotImplementedError
|
@@ -239,8 +223,12 @@ class KoBEST(datasets.GeneratorBasedBuilder):
|
|
239 |
|
240 |
|
241 |
if __name__ == "__main__":
|
242 |
-
|
243 |
-
|
244 |
-
|
245 |
-
|
|
|
|
|
|
|
|
|
246 |
|
|
|
34 |
"train": _URL + "/v1.0/SentiNeg/train.tsv",
|
35 |
"dev": _URL + "/v1.0/SentiNeg/dev.tsv",
|
36 |
"test": _URL + "/v1.0/SentiNeg/test.tsv",
|
37 |
+
"test_originated": _URL + "/v1.0/SentiNeg/test.tsv",
|
38 |
},
|
39 |
"hellaswag": {
|
40 |
"train": _URL + "/v1.0/HellaSwag/train.tsv",
|
|
|
70 |
self.url = url
|
71 |
|
72 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
73 |
class KoBEST(datasets.GeneratorBasedBuilder):
|
74 |
BUILDER_CONFIGS = [
|
75 |
KoBESTConfig(name=name, description=_DESCRIPTION, data_url=_DATA_URLS[name], citation=_CITATAION, url=_URL)
|
|
|
125 |
dev = dl_manager.download_and_extract(self.config.data_url["dev"])
|
126 |
test = dl_manager.download_and_extract(self.config.data_url["test"])
|
127 |
|
128 |
+
if self.config.data_url.get("test_originated"):
|
129 |
+
test_originated = dl_manager.download_and_extract(self.config.data_url["test_originated"])
|
130 |
+
|
131 |
+
return [
|
132 |
+
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": train, "split": "train"}),
|
133 |
+
datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": dev, "split": "dev"}),
|
134 |
+
datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": test, "split": "test"}),
|
135 |
+
datasets.SplitGenerator(name="test_originated", gen_kwargs={"filepath": test_originated, "split": "test_originated"}),
|
136 |
+
]
|
137 |
+
|
138 |
return [
|
139 |
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": train, "split": "train"}),
|
140 |
datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": dev, "split": "dev"}),
|
|
|
186 |
df = df.dropna()
|
187 |
df = df[['context', 'choice1', 'choice2', 'choice3', 'choice4', 'label']]
|
188 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
189 |
df = df.rename(columns={
|
190 |
'context': 'context',
|
191 |
'choice1': 'ending_1',
|
|
|
198 |
elif self.config.name == "sentineg":
|
199 |
df = pd.read_csv(filepath, sep="\t")
|
200 |
df = df.dropna()
|
|
|
201 |
|
202 |
+
if split == "test_originated":
|
203 |
+
df = df[['Text_origin', 'Label_origin']]
|
|
|
|
|
|
|
204 |
|
205 |
+
df = df.rename(columns={
|
206 |
+
'Text_origin': 'sentence',
|
207 |
+
'Label_origin': 'label',
|
208 |
+
})
|
209 |
+
else:
|
210 |
+
df = df[['Text', 'Label']]
|
211 |
+
|
212 |
+
df = df.rename(columns={
|
213 |
+
'Text': 'sentence',
|
214 |
+
'Label': 'label',
|
215 |
+
})
|
216 |
|
217 |
else:
|
218 |
raise NotImplementedError
|
|
|
223 |
|
224 |
|
225 |
if __name__ == "__main__":
|
226 |
+
dataset = datasets.load_dataset("kobest_v1.py", 'sentineg', ignore_verifications=True)
|
227 |
+
ds = dataset['test_originated']
|
228 |
+
print(ds)
|
229 |
+
|
230 |
+
# for task in ['boolq', 'copa', 'wic', 'hellaswag', 'sentineg']:
|
231 |
+
# dataset = datasets.load_dataset("kobest_v1.py", task, ignore_verifications=True)
|
232 |
+
# print(dataset)
|
233 |
+
# print(dataset['train']['label'])
|
234 |
|