metadata
annotations_creators:
- crowdsourced
language:
- en
language_creators:
- found
license:
- unknown
multilinguality:
- monolingual
pretty_name: Abductive Reasoning in narrative Text
size_categories:
- 100K<n<1M
source_datasets:
- original
task_categories:
- multiple-choice
- text-classification
task_ids:
- natural-language-inference
paperswithcode_id: art-dataset
tags:
- abductive-natural-language-inference
dataset_info:
features:
- name: observation_1
dtype: string
- name: observation_2
dtype: string
- name: hypothesis_1
dtype: string
- name: hypothesis_2
dtype: string
- name: label
dtype:
class_label:
names:
'0': '0'
'1': '1'
'2': '2'
config_name: anli
splits:
- name: validation
num_bytes: 312314
num_examples: 1532
- name: train
num_bytes: 34046304
num_examples: 169654
download_size: 5118294
dataset_size: 34358618
Dataset Card for "art"
Table of Contents
- Dataset Description
- Dataset Structure
- Dataset Creation
- Considerations for Using the Data
- Additional Information
Dataset Description
- Homepage: https://leaderboard.allenai.org/anli/submissions/get-started
- Repository: https://github.com/allenai/abductive-commonsense-reasoning
- Paper: Abductive Commonsense Reasoning
- Point of Contact: More Information Needed
- Size of downloaded dataset files: 4.88 MB
- Size of the generated dataset: 32.77 MB
- Total amount of disk used: 37.65 MB
Dataset Summary
ART consists of over 20k commonsense narrative contexts and 200k explanations.
The Abductive Natural Language Inference Dataset from AI2.
Supported Tasks and Leaderboards
Languages
Dataset Structure
Data Instances
anli
- Size of downloaded dataset files: 4.88 MB
- Size of the generated dataset: 32.77 MB
- Total amount of disk used: 37.65 MB
An example of 'train' looks as follows.
{
"hypothesis_1": "Chad's car had all sorts of other problems besides alignment.",
"hypothesis_2": "Chad's car had all sorts of benefits other than being sexy.",
"label": 1,
"observation_1": "Chad went to get the wheel alignment measured on his car.",
"observation_2": "The mechanic provided a working alignment with new body work."
}
Data Fields
The data fields are the same among all splits.
anli
observation_1: astringfeature.observation_2: astringfeature.hypothesis_1: astringfeature.hypothesis_2: astringfeature.label: a classification label, with possible values including0(0),1(1),2(2).
Data Splits
| name | train | validation |
|---|---|---|
| anli | 169654 | 1532 |
Dataset Creation
Curation Rationale
Source Data
Initial Data Collection and Normalization
Who are the source language producers?
Annotations
Annotation process
Who are the annotators?
Personal and Sensitive Information
Considerations for Using the Data
Social Impact of Dataset
Discussion of Biases
Other Known Limitations
Additional Information
Dataset Curators
Licensing Information
Citation Information
@inproceedings{Bhagavatula2020Abductive,
title={Abductive Commonsense Reasoning},
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},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=Byg1v1HKDB}
}
Contributions
Thanks to @patrickvonplaten, @thomwolf, @mariamabarham, @lewtun, @lhoestq for adding this dataset.