Datasets:
Tasks:
Text Classification
Modalities:
Text
Sub-tasks:
multi-class-classification
Languages:
English
Size:
10K - 100K
ArXiv:
License:
Remove Medical Subdomain of Clinical Notes
Browse files
data/medical_subdomain_of_clinical_notes/task.json
DELETED
@@ -1 +0,0 @@
|
|
1 |
-
{"name": "medical_subdomain_of_clinical_notes", "description": "", "data_columns": ["Note", "ID"], "label_columns": {"Label": ["cardiology", "gastroenterology", "nephrology", "neurology", "psychiatry", "pulmonary disease"]}}
|
|
|
|
data/medical_subdomain_of_clinical_notes/test_unlabeled.csv
DELETED
The diff for this file is too large to render.
See raw diff
|
|
data/medical_subdomain_of_clinical_notes/train.csv
DELETED
The diff for this file is too large to render.
See raw diff
|
|
raft.py
CHANGED
@@ -21,7 +21,6 @@ from pathlib import Path
|
|
21 |
import datasets
|
22 |
|
23 |
# TODO: Add BibTeX citation
|
24 |
-
# Find for instance the citation on arxiv or on the dataset repo/website
|
25 |
_CITATION = """\
|
26 |
@InProceedings{huggingface:dataset,
|
27 |
title = {A great new dataset},
|
@@ -31,44 +30,32 @@ year={2020}
|
|
31 |
}
|
32 |
"""
|
33 |
|
34 |
-
|
35 |
-
|
|
|
|
|
|
|
|
|
|
|
36 |
"""
|
37 |
|
38 |
-
|
39 |
-
_HOMEPAGE = ""
|
40 |
|
41 |
# TODO: Add the licence for the dataset here if you can find it
|
42 |
_LICENSE = ""
|
43 |
|
44 |
-
|
45 |
-
# The HuggingFace dataset library don't host the datasets but only point to the original files
|
46 |
-
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
47 |
-
# This gets all folders within the directory named `data`
|
48 |
-
DATA_DIR_URL = "data/" # "https://huggingface.co/datasets/ought/raft/resolve/main/data/"
|
49 |
-
# print([p for p in DATA_DIR_PATH.iterdir() if p.is_dir()])
|
50 |
TASKS = {
|
51 |
"ade_corpus_v2": {
|
52 |
"name": "ade_corpus_v2",
|
53 |
"description": "",
|
54 |
-
"data_columns": [
|
55 |
-
|
56 |
-
"ID"
|
57 |
-
],
|
58 |
-
"label_columns": {
|
59 |
-
"Label": [
|
60 |
-
"ADE-related",
|
61 |
-
"not ADE-related"
|
62 |
-
]
|
63 |
-
}
|
64 |
},
|
65 |
"banking_77": {
|
66 |
"name": "banking_77",
|
67 |
"description": "",
|
68 |
-
"data_columns": [
|
69 |
-
"Query",
|
70 |
-
"ID"
|
71 |
-
],
|
72 |
"label_columns": {
|
73 |
"Label": [
|
74 |
"Refund_not_showing_up",
|
@@ -147,23 +134,15 @@ TASKS = {
|
|
147 |
"visa_or_mastercard",
|
148 |
"why_verify_identity",
|
149 |
"wrong_amount_of_cash_received",
|
150 |
-
"wrong_exchange_rate_for_cash_withdrawal"
|
151 |
]
|
152 |
-
}
|
153 |
},
|
154 |
"terms_of_service": {
|
155 |
"name": "terms_of_service",
|
156 |
"description": "",
|
157 |
-
"data_columns": [
|
158 |
-
|
159 |
-
"ID"
|
160 |
-
],
|
161 |
-
"label_columns": {
|
162 |
-
"Label": [
|
163 |
-
"not potentially unfair",
|
164 |
-
"potentially unfair"
|
165 |
-
]
|
166 |
-
}
|
167 |
},
|
168 |
"tai_safety_research": {
|
169 |
"name": "tai_safety_research",
|
@@ -176,138 +155,51 @@ TASKS = {
|
|
176 |
"Item Type",
|
177 |
"Author",
|
178 |
"Publication Title",
|
179 |
-
"ID"
|
180 |
],
|
181 |
-
"label_columns": {
|
182 |
-
"Label": [
|
183 |
-
"TAI safety research",
|
184 |
-
"not TAI safety research"
|
185 |
-
]
|
186 |
-
}
|
187 |
},
|
188 |
"neurips_impact_statement_risks": {
|
189 |
"name": "neurips_impact_statement_risks",
|
190 |
"description": "",
|
191 |
-
"data_columns": [
|
192 |
-
|
193 |
-
"Paper link",
|
194 |
-
"Impact statement",
|
195 |
-
"ID"
|
196 |
-
],
|
197 |
-
"label_columns": {
|
198 |
-
"Label": [
|
199 |
-
"doesn't mention a harmful application",
|
200 |
-
"mentions a harmful application"
|
201 |
-
]
|
202 |
-
}
|
203 |
-
},
|
204 |
-
"medical_subdomain_of_clinical_notes": {
|
205 |
-
"name": "medical_subdomain_of_clinical_notes",
|
206 |
-
"description": "",
|
207 |
-
"data_columns": [
|
208 |
-
"Note",
|
209 |
-
"ID"
|
210 |
-
],
|
211 |
-
"label_columns": {
|
212 |
-
"Label": [
|
213 |
-
"cardiology",
|
214 |
-
"gastroenterology",
|
215 |
-
"nephrology",
|
216 |
-
"neurology",
|
217 |
-
"psychiatry",
|
218 |
-
"pulmonary disease"
|
219 |
-
]
|
220 |
-
}
|
221 |
},
|
222 |
"overruling": {
|
223 |
"name": "overruling",
|
224 |
"description": "",
|
225 |
-
"data_columns": [
|
226 |
-
|
227 |
-
"ID"
|
228 |
-
],
|
229 |
-
"label_columns": {
|
230 |
-
"Label": [
|
231 |
-
"not overruling",
|
232 |
-
"overruling"
|
233 |
-
]
|
234 |
-
}
|
235 |
},
|
236 |
"systematic_review_inclusion": {
|
237 |
"name": "systematic_review_inclusion",
|
238 |
"description": "",
|
239 |
-
"data_columns": [
|
240 |
-
|
241 |
-
"Abstract",
|
242 |
-
"Authors",
|
243 |
-
"Journal",
|
244 |
-
"ID"
|
245 |
-
],
|
246 |
-
"label_columns": {
|
247 |
-
"Label": [
|
248 |
-
"included",
|
249 |
-
"not included"
|
250 |
-
]
|
251 |
-
}
|
252 |
},
|
253 |
"one_stop_english": {
|
254 |
"name": "one_stop_english",
|
255 |
"description": "",
|
256 |
-
"data_columns": [
|
257 |
-
|
258 |
-
"ID"
|
259 |
-
],
|
260 |
-
"label_columns": {
|
261 |
-
"Label": [
|
262 |
-
"advanced",
|
263 |
-
"elementary",
|
264 |
-
"intermediate"
|
265 |
-
]
|
266 |
-
}
|
267 |
},
|
268 |
"tweet_eval_hate": {
|
269 |
"name": "tweet_eval_hate",
|
270 |
"description": "",
|
271 |
-
"data_columns": [
|
272 |
-
|
273 |
-
"ID"
|
274 |
-
],
|
275 |
-
"label_columns": {
|
276 |
-
"Label": [
|
277 |
-
"hate speech",
|
278 |
-
"not hate speech"
|
279 |
-
]
|
280 |
-
}
|
281 |
},
|
282 |
"twitter_complaints": {
|
283 |
"name": "twitter_complaints",
|
284 |
"description": "",
|
285 |
-
"data_columns": [
|
286 |
-
|
287 |
-
"ID"
|
288 |
-
],
|
289 |
-
"label_columns": {
|
290 |
-
"Label": [
|
291 |
-
"complaint",
|
292 |
-
"no complaint"
|
293 |
-
]
|
294 |
-
}
|
295 |
},
|
296 |
"semiconductor_org_types": {
|
297 |
"name": "semiconductor_org_types",
|
298 |
"description": "",
|
299 |
-
"data_columns": [
|
300 |
-
|
301 |
-
"Organization name",
|
302 |
-
"ID"
|
303 |
-
],
|
304 |
-
"label_columns": {
|
305 |
-
"Label": [
|
306 |
-
"company",
|
307 |
-
"research institute",
|
308 |
-
"university"
|
309 |
-
]
|
310 |
-
}
|
311 |
},
|
312 |
}
|
313 |
|
|
|
21 |
import datasets
|
22 |
|
23 |
# TODO: Add BibTeX citation
|
|
|
24 |
_CITATION = """\
|
25 |
@InProceedings{huggingface:dataset,
|
26 |
title = {A great new dataset},
|
|
|
30 |
}
|
31 |
"""
|
32 |
|
33 |
+
_DESCRIPTION = """Large pre-trained language models have shown promise for few-shot learning, completing text-based tasks given only a few task-specific examples. Will models soon solve classification tasks that have so far been reserved for human research assistants?
|
34 |
+
|
35 |
+
[RAFT](https://raft.elicit.org) is a few-shot classification benchmark that tests language models:
|
36 |
+
|
37 |
+
- across multiple domains (lit review, tweets, customer interaction, etc.)
|
38 |
+
- on economically valuable classification tasks (someone inherently cares about the task)
|
39 |
+
- in a setting that mirrors deployment (50 examples per task, info retrieval allowed, hidden test set)
|
40 |
"""
|
41 |
|
42 |
+
_HOMEPAGE = "https://raft.elicit.org"
|
|
|
43 |
|
44 |
# TODO: Add the licence for the dataset here if you can find it
|
45 |
_LICENSE = ""
|
46 |
|
47 |
+
DATA_DIR_URL = "data/"
|
|
|
|
|
|
|
|
|
|
|
48 |
TASKS = {
|
49 |
"ade_corpus_v2": {
|
50 |
"name": "ade_corpus_v2",
|
51 |
"description": "",
|
52 |
+
"data_columns": ["Sentence", "ID"],
|
53 |
+
"label_columns": {"Label": ["ADE-related", "not ADE-related"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
},
|
55 |
"banking_77": {
|
56 |
"name": "banking_77",
|
57 |
"description": "",
|
58 |
+
"data_columns": ["Query", "ID"],
|
|
|
|
|
|
|
59 |
"label_columns": {
|
60 |
"Label": [
|
61 |
"Refund_not_showing_up",
|
|
|
134 |
"visa_or_mastercard",
|
135 |
"why_verify_identity",
|
136 |
"wrong_amount_of_cash_received",
|
137 |
+
"wrong_exchange_rate_for_cash_withdrawal",
|
138 |
]
|
139 |
+
},
|
140 |
},
|
141 |
"terms_of_service": {
|
142 |
"name": "terms_of_service",
|
143 |
"description": "",
|
144 |
+
"data_columns": ["Sentence", "ID"],
|
145 |
+
"label_columns": {"Label": ["not potentially unfair", "potentially unfair"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
146 |
},
|
147 |
"tai_safety_research": {
|
148 |
"name": "tai_safety_research",
|
|
|
155 |
"Item Type",
|
156 |
"Author",
|
157 |
"Publication Title",
|
158 |
+
"ID",
|
159 |
],
|
160 |
+
"label_columns": {"Label": ["TAI safety research", "not TAI safety research"]},
|
|
|
|
|
|
|
|
|
|
|
161 |
},
|
162 |
"neurips_impact_statement_risks": {
|
163 |
"name": "neurips_impact_statement_risks",
|
164 |
"description": "",
|
165 |
+
"data_columns": ["Paper title", "Paper link", "Impact statement", "ID"],
|
166 |
+
"label_columns": {"Label": ["doesn't mention a harmful application", "mentions a harmful application"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
167 |
},
|
168 |
"overruling": {
|
169 |
"name": "overruling",
|
170 |
"description": "",
|
171 |
+
"data_columns": ["Sentence", "ID"],
|
172 |
+
"label_columns": {"Label": ["not overruling", "overruling"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
173 |
},
|
174 |
"systematic_review_inclusion": {
|
175 |
"name": "systematic_review_inclusion",
|
176 |
"description": "",
|
177 |
+
"data_columns": ["Title", "Abstract", "Authors", "Journal", "ID"],
|
178 |
+
"label_columns": {"Label": ["included", "not included"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
179 |
},
|
180 |
"one_stop_english": {
|
181 |
"name": "one_stop_english",
|
182 |
"description": "",
|
183 |
+
"data_columns": ["Article", "ID"],
|
184 |
+
"label_columns": {"Label": ["advanced", "elementary", "intermediate"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
185 |
},
|
186 |
"tweet_eval_hate": {
|
187 |
"name": "tweet_eval_hate",
|
188 |
"description": "",
|
189 |
+
"data_columns": ["Tweet", "ID"],
|
190 |
+
"label_columns": {"Label": ["hate speech", "not hate speech"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
191 |
},
|
192 |
"twitter_complaints": {
|
193 |
"name": "twitter_complaints",
|
194 |
"description": "",
|
195 |
+
"data_columns": ["Tweet text", "ID"],
|
196 |
+
"label_columns": {"Label": ["complaint", "no complaint"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
197 |
},
|
198 |
"semiconductor_org_types": {
|
199 |
"name": "semiconductor_org_types",
|
200 |
"description": "",
|
201 |
+
"data_columns": ["Paper title", "Organization name", "ID"],
|
202 |
+
"label_columns": {"Label": ["company", "research institute", "university"]},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
203 |
},
|
204 |
}
|
205 |
|