Datasets:
GEM
/

Tasks:
Other
Modalities:
Text
Languages:
English
ArXiv:
Libraries:
Datasets
License:
Sebastian Gehrmann commited on
Commit
c022215
1 Parent(s): 7d5b162

data card.

Browse files
Files changed (1) hide show
  1. README.md +481 -0
README.md CHANGED
@@ -0,0 +1,481 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ annotations_creators:
3
+ - automatically-created
4
+ language_creators:
5
+ - unknown
6
+ languages:
7
+ - unknown
8
+ licenses:
9
+ - apache-2.0
10
+ multilinguality:
11
+ - unknown
12
+ pretty_name: ART
13
+ size_categories:
14
+ - unknown
15
+ source_datasets:
16
+ - original
17
+ task_categories:
18
+ - reasoning
19
+ task_ids:
20
+ - unknown
21
+ ---
22
+
23
+ # Dataset Card for GEM/ART
24
+
25
+ ## Dataset Description
26
+
27
+ - **Homepage:** http://abductivecommonsense.xyz/
28
+ - **Repository:** https://storage.googleapis.com/ai2-mosaic/public/abductive-commonsense-reasoning-iclr2020/anlg.zip
29
+ - **Paper:** https://openreview.net/pdf?id=Byg1v1HKDB
30
+ - **Leaderboard:** N/A
31
+ - **Point of Contact:** Chandra Bhagavatulla
32
+
33
+ ### Link to Main Data Card
34
+
35
+ You can find the main data card on the [GEM Website](https://gem-benchmark.com/data_cards/ART).
36
+
37
+ ### Dataset Summary
38
+
39
+ Abductive reasoning is inference to the most plausible explanation. For example, if Jenny finds her house in a mess when she returns from work, and remembers that she left a window open, she can hypothesize that a thief broke into her house and caused the mess, as the most plausible explanation.
40
+ This data loader focuses on abductive NLG: a conditional English generation task for explaining given observations in natural language.
41
+
42
+ You can load the dataset via:
43
+ ```
44
+ import datasets
45
+ data = datasets.load_dataset('GEM/ART')
46
+ ```
47
+ The data loader can be found [here](https://huggingface.co/datasets/GEM/ART).
48
+
49
+ #### website
50
+ [Website](http://abductivecommonsense.xyz/)
51
+
52
+ #### paper
53
+ [OpenReview](https://openreview.net/pdf?id=Byg1v1HKDB)
54
+
55
+ #### authors
56
+ Chandra Bhagavatula (AI2), Ronan Le Bras (AI2), Chaitanya Malaviya (AI2), Keisuke Sakaguchi (AI2), Ari Holtzman (AI2, UW), Hannah Rashkin (AI2, UW), Doug Downey (AI2), Wen-tau Yih (AI2), Yejin Choi (AI2, UW)
57
+
58
+ ## Dataset Overview
59
+
60
+ ### Where to find the Data and its Documentation
61
+
62
+ #### Webpage
63
+
64
+ <!-- info: What is the webpage for the dataset (if it exists)? -->
65
+ <!-- scope: telescope -->
66
+ [Website](http://abductivecommonsense.xyz/)
67
+
68
+ #### Download
69
+
70
+ <!-- info: What is the link to where the original dataset is hosted? -->
71
+ <!-- scope: telescope -->
72
+ [Google Storage](https://storage.googleapis.com/ai2-mosaic/public/abductive-commonsense-reasoning-iclr2020/anlg.zip)
73
+
74
+ #### Paper
75
+
76
+ <!-- info: What is the link to the paper describing the dataset (open access preferred)? -->
77
+ <!-- scope: telescope -->
78
+ [OpenReview](https://openreview.net/pdf?id=Byg1v1HKDB)
79
+
80
+ #### BibTex
81
+
82
+ <!-- info: Provide the BibTex-formatted reference for the dataset. Please use the correct published version (ACL anthology, etc.) instead of google scholar created Bibtex. -->
83
+ <!-- scope: microscope -->
84
+ ```
85
+ @inproceedings{
86
+ Bhagavatula2020Abductive,
87
+ title={Abductive Commonsense Reasoning},
88
+ author={Chandra Bhagavatula and Ronan Le Bras and Chaitanya Malaviya and Keisuke Sakaguchi and Ari Holtzman and Hannah Rashkin and Doug Downey and Wen-tau Yih and Yejin Choi},
89
+ booktitle={International Conference on Learning Representations},
90
+ year={2020},
91
+ url={https://openreview.net/forum?id=Byg1v1HKDB}
92
+ }
93
+ ```
94
+
95
+ #### Contact Name
96
+
97
+ <!-- quick -->
98
+ <!-- info: If known, provide the name of at least one person the reader can contact for questions about the dataset. -->
99
+ <!-- scope: periscope -->
100
+ Chandra Bhagavatulla
101
+
102
+ #### Contact Email
103
+
104
+ <!-- info: If known, provide the email of at least one person the reader can contact for questions about the dataset. -->
105
+ <!-- scope: periscope -->
106
107
+
108
+ #### Has a Leaderboard?
109
+
110
+ <!-- info: Does the dataset have an active leaderboard? -->
111
+ <!-- scope: telescope -->
112
+ no
113
+
114
+
115
+ ### Languages and Intended Use
116
+
117
+ #### Multilingual?
118
+
119
+ <!-- quick -->
120
+ <!-- info: Is the dataset multilingual? -->
121
+ <!-- scope: telescope -->
122
+ no
123
+
124
+ #### Covered Languages
125
+
126
+ <!-- quick -->
127
+ <!-- info: What languages/dialects are covered in the dataset? -->
128
+ <!-- scope: telescope -->
129
+ `English`
130
+
131
+ #### Whose Language?
132
+
133
+ <!-- info: Whose language is in the dataset? -->
134
+ <!-- scope: periscope -->
135
+ Crowdworkers on the Amazon Mechanical Turk platform based in the U.S, Canada, U.K and Australia.
136
+
137
+ #### License
138
+
139
+ <!-- quick -->
140
+ <!-- info: What is the license of the dataset? -->
141
+ <!-- scope: telescope -->
142
+ apache-2.0: Apache License 2.0
143
+
144
+ #### Intended Use
145
+
146
+ <!-- info: What is the intended use of the dataset? -->
147
+ <!-- scope: microscope -->
148
+ To study the viability of language-based abductive reasoning. Training and evaluating models to generate a plausible hypothesis to explain two given observations.
149
+
150
+ #### Primary Task
151
+
152
+ <!-- info: What primary task does the dataset support? -->
153
+ <!-- scope: telescope -->
154
+ Reasoning
155
+
156
+
157
+ ### Credit
158
+
159
+ #### Curation Organization Type(s)
160
+
161
+ <!-- info: In what kind of organization did the dataset curation happen? -->
162
+ <!-- scope: telescope -->
163
+ `industry`
164
+
165
+ #### Curation Organization(s)
166
+
167
+ <!-- info: Name the organization(s). -->
168
+ <!-- scope: periscope -->
169
+ Allen Institute for AI
170
+
171
+ #### Dataset Creators
172
+
173
+ <!-- info: Who created the original dataset? List the people involved in collecting the dataset and their affiliation(s). -->
174
+ <!-- scope: microscope -->
175
+ Chandra Bhagavatula (AI2), Ronan Le Bras (AI2), Chaitanya Malaviya (AI2), Keisuke Sakaguchi (AI2), Ari Holtzman (AI2, UW), Hannah Rashkin (AI2, UW), Doug Downey (AI2), Wen-tau Yih (AI2), Yejin Choi (AI2, UW)
176
+
177
+ #### Funding
178
+
179
+ <!-- info: Who funded the data creation? -->
180
+ <!-- scope: microscope -->
181
+ Allen Institute for AI
182
+
183
+ #### Who added the Dataset to GEM?
184
+
185
+ <!-- info: Who contributed to the data card and adding the dataset to GEM? List the people+affiliations involved in creating this data card and who helped integrate this dataset into GEM. -->
186
+ <!-- scope: microscope -->
187
+ Chandra Bhagavatula (AI2), Ronan LeBras (AI2), Aman Madaan (CMU), Nico Daheim (RWTH Aachen University)
188
+
189
+
190
+ ### Dataset Structure
191
+
192
+ #### Data Fields
193
+
194
+ <!-- info: List and describe the fields present in the dataset. -->
195
+ <!-- scope: telescope -->
196
+ - `observation_1`: A string describing an observation / event.
197
+ - `observation_2`: A string describing an observation / event.
198
+ - `label`: A string that plausibly explains why observation_1 and observation_2 might have happened.
199
+
200
+ #### How were labels chosen?
201
+
202
+ <!-- info: How were the labels chosen? -->
203
+ <!-- scope: microscope -->
204
+ Explanations were authored by crowdworkers on the Amazon Mechanical Turk platform using a custom template designed by the creators of the dataset.
205
+
206
+ #### Example Instance
207
+
208
+ <!-- info: Provide a JSON formatted example of a typical instance in the dataset. -->
209
+ <!-- scope: periscope -->
210
+ ```
211
+ {
212
+ 'gem_id': 'GEM-ART-validation-0',
213
+ 'observation_1': 'Stephen was at a party.',
214
+ 'observation_2': 'He checked it but it was completely broken.',
215
+ 'label': 'Stephen knocked over a vase while drunk.'
216
+ }
217
+ ```
218
+
219
+ #### Data Splits
220
+
221
+ <!-- info: Describe and name the splits in the dataset if there are more than one. -->
222
+ <!-- scope: periscope -->
223
+ - `train`: Consists of training instances.
224
+ - `dev`: Consists of dev instances.
225
+ - `test`: Consists of test instances.
226
+
227
+
228
+
229
+
230
+ ## Dataset in GEM
231
+
232
+ ### Rationale for Inclusion in GEM
233
+
234
+ #### Why is the Dataset in GEM?
235
+
236
+ <!-- info: What does this dataset contribute toward better generation evaluation and why is it part of GEM? -->
237
+ <!-- scope: microscope -->
238
+ Abductive reasoning is a crucial capability of humans and ART is the first dataset curated to study language-based abductive reasoning.
239
+
240
+ #### Similar Datasets
241
+
242
+ <!-- info: Do other datasets for the high level task exist? -->
243
+ <!-- scope: telescope -->
244
+ no
245
+
246
+ #### Ability that the Dataset measures
247
+
248
+ <!-- info: What aspect of model ability can be measured with this dataset? -->
249
+ <!-- scope: periscope -->
250
+ Whether models can reason abductively about a given pair of observations.
251
+
252
+
253
+ ### GEM-Specific Curation
254
+
255
+ #### Modificatied for GEM?
256
+
257
+ <!-- info: Has the GEM version of the dataset been modified in any way (data, processing, splits) from the original curated data? -->
258
+ <!-- scope: telescope -->
259
+ no
260
+
261
+ #### Additional Splits?
262
+
263
+ <!-- info: Does GEM provide additional splits to the dataset? -->
264
+ <!-- scope: telescope -->
265
+ no
266
+
267
+
268
+ ### Getting Started with the Task
269
+
270
+ #### Pointers to Resources
271
+
272
+ <!-- info: Getting started with in-depth research on the task. Add relevant pointers to resources that researchers can consult when they want to get started digging deeper into the task. -->
273
+ <!-- scope: microscope -->
274
+ - Paper: https://arxiv.org/abs/1908.05739
275
+ - Code: https://github.com/allenai/abductive-commonsense-reasoning
276
+
277
+
278
+
279
+ ## Previous Results
280
+
281
+ ### Previous Results
282
+
283
+ #### Measured Model Abilities
284
+
285
+ <!-- info: What aspect of model ability can be measured with this dataset? -->
286
+ <!-- scope: telescope -->
287
+ Whether models can reason abductively about a given pair of observations.
288
+
289
+ #### Metrics
290
+
291
+ <!-- info: What metrics are typically used for this task? -->
292
+ <!-- scope: periscope -->
293
+ `BLEU`, `BERT-Score`, `ROUGE`
294
+
295
+ #### Previous results available?
296
+
297
+ <!-- info: Are previous results available? -->
298
+ <!-- scope: telescope -->
299
+ no
300
+
301
+
302
+
303
+ ## Dataset Curation
304
+
305
+ ### Original Curation
306
+
307
+ #### Sourced from Different Sources
308
+
309
+ <!-- info: Is the dataset aggregated from different data sources? -->
310
+ <!-- scope: telescope -->
311
+ no
312
+
313
+
314
+ ### Language Data
315
+
316
+ #### How was Language Data Obtained?
317
+
318
+ <!-- info: How was the language data obtained? -->
319
+ <!-- scope: telescope -->
320
+ `Crowdsourced`
321
+
322
+ #### Where was it crowdsourced?
323
+
324
+ <!-- info: If crowdsourced, where from? -->
325
+ <!-- scope: periscope -->
326
+ `Amazon Mechanical Turk`
327
+
328
+ #### Language Producers
329
+
330
+ <!-- info: What further information do we have on the language producers? -->
331
+ <!-- scope: microscope -->
332
+ Language producers were English speakers in U.S., Canada, U.K and Australia.
333
+
334
+ #### Topics Covered
335
+
336
+ <!-- info: Does the language in the dataset focus on specific topics? How would you describe them? -->
337
+ <!-- scope: periscope -->
338
+ No
339
+
340
+ #### Data Validation
341
+
342
+ <!-- info: Was the text validated by a different worker or a data curator? -->
343
+ <!-- scope: telescope -->
344
+ validated by crowdworker
345
+
346
+ #### Was Data Filtered?
347
+
348
+ <!-- info: Were text instances selected or filtered? -->
349
+ <!-- scope: telescope -->
350
+ algorithmically
351
+
352
+ #### Filter Criteria
353
+
354
+ <!-- info: What were the selection criteria? -->
355
+ <!-- scope: microscope -->
356
+ Adversarial filtering algorithm as described in the paper: https://arxiv.org/abs/1908.05739
357
+
358
+
359
+ ### Structured Annotations
360
+
361
+ #### Additional Annotations?
362
+
363
+ <!-- quick -->
364
+ <!-- info: Does the dataset have additional annotations for each instance? -->
365
+ <!-- scope: telescope -->
366
+ automatically created
367
+
368
+ #### Annotation Service?
369
+
370
+ <!-- info: Was an annotation service used? -->
371
+ <!-- scope: telescope -->
372
+ no
373
+
374
+ #### Annotation Values
375
+
376
+ <!-- info: Purpose and values for each annotation -->
377
+ <!-- scope: microscope -->
378
+ Each observation is associated with a list of COMET (https://arxiv.org/abs/1906.05317) inferences.
379
+
380
+ #### Any Quality Control?
381
+
382
+ <!-- info: Quality control measures? -->
383
+ <!-- scope: telescope -->
384
+ none
385
+
386
+
387
+ ### Consent
388
+
389
+ #### Any Consent Policy?
390
+
391
+ <!-- info: Was there a consent policy involved when gathering the data? -->
392
+ <!-- scope: telescope -->
393
+ no
394
+
395
+
396
+ ### Private Identifying Information (PII)
397
+
398
+ #### Contains PII?
399
+
400
+ <!-- quick -->
401
+ <!-- info: Does the source language data likely contain Personal Identifying Information about the data creators or subjects? -->
402
+ <!-- scope: telescope -->
403
+ no PII
404
+
405
+ #### Justification for no PII
406
+
407
+ <!-- info: Provide a justification for selecting `no PII` above. -->
408
+ <!-- scope: periscope -->
409
+ The dataset contains day-to-day events. It does not contain names, emails, addresses etc.
410
+
411
+
412
+ ### Maintenance
413
+
414
+ #### Any Maintenance Plan?
415
+
416
+ <!-- info: Does the original dataset have a maintenance plan? -->
417
+ <!-- scope: telescope -->
418
+ no
419
+
420
+
421
+
422
+ ## Broader Social Context
423
+
424
+ ### Previous Work on the Social Impact of the Dataset
425
+
426
+ #### Usage of Models based on the Data
427
+
428
+ <!-- info: Are you aware of cases where models trained on the task featured in this dataset ore related tasks have been used in automated systems? -->
429
+ <!-- scope: telescope -->
430
+ no
431
+
432
+
433
+ ### Impact on Under-Served Communities
434
+
435
+ #### Addresses needs of underserved Communities?
436
+
437
+ <!-- info: Does this dataset address the needs of communities that are traditionally underserved in language technology, and particularly language generation technology? Communities may be underserved for exemple because their language, language variety, or social or geographical context is underepresented in NLP and NLG resources (datasets and models). -->
438
+ <!-- scope: telescope -->
439
+ no
440
+
441
+
442
+ ### Discussion of Biases
443
+
444
+ #### Any Documented Social Biases?
445
+
446
+ <!-- info: Are there documented social biases in the dataset? Biases in this context are variations in the ways members of different social categories are represented that can have harmful downstream consequences for members of the more disadvantaged group. -->
447
+ <!-- scope: telescope -->
448
+ no
449
+
450
+
451
+
452
+ ## Considerations for Using the Data
453
+
454
+ ### PII Risks and Liability
455
+
456
+ #### Potential PII Risk
457
+
458
+ <!-- info: Considering your answers to the PII part of the Data Curation Section, describe any potential privacy to the data subjects and creators risks when using the dataset. -->
459
+ <!-- scope: microscope -->
460
+ None
461
+
462
+
463
+ ### Licenses
464
+
465
+ #### Copyright Restrictions on the Dataset
466
+
467
+ <!-- info: Based on your answers in the Intended Use part of the Data Overview Section, which of the following best describe the copyright and licensing status of the dataset? -->
468
+ <!-- scope: periscope -->
469
+ `public domain`
470
+
471
+ #### Copyright Restrictions on the Language Data
472
+
473
+ <!-- info: Based on your answers in the Language part of the Data Curation Section, which of the following best describe the copyright and licensing status of the underlying language data? -->
474
+ <!-- scope: periscope -->
475
+ `public domain`
476
+
477
+
478
+ ### Known Technical Limitations
479
+
480
+
481
+