Datasets:
GEM
/

Tasks:
Other
Modalities:
Text
Languages:
English
ArXiv:
Libraries:
Datasets
License:
Sebastian Gehrmann commited on
Commit
d89ac97
1 Parent(s): 18cbc16

data card.

Browse files
Files changed (2) hide show
  1. FairytaleQA.json +5 -5
  2. README.md +574 -94
FairytaleQA.json CHANGED
@@ -5,10 +5,10 @@
5
  },
6
  "where": {
7
  "has-leaderboard": "yes",
8
- "leaderboard-url": "https://paperswithcode.com/sota/question-generation-on-fairytaleqa",
9
  "leaderboard-description": "The task was to generate questions corresponding to the given answers and the story context. Success on the Question Generation task is typically measured by achieving a high ROUGE-L score to the reference ground-truth question.",
10
- "data-url": "https://github.com/uci-soe/FairytaleQAData",
11
- "paper-url": "https://arxiv.org/abs/2203.13947",
12
  "paper-bibtext": "@inproceedings{xu2022fairytaleqa,\n author={Xu, Ying and Wang, Dakuo and Yu, Mo and Ritchie, Daniel and Yao, Bingsheng and Wu, Tongshuang and Zhang, Zheng and Li, Toby Jia-Jun and Bradford, Nora and Sun, Branda and Hoang, Tran Bao and Sang, Yisi and Hou, Yufang and Ma, Xiaojuan and Yang, Diyi and Peng, Nanyun and Yu, Zhou and Warschauer, Mark},\n title = {Fantastic Questions and Where to Find Them: Fairytale{QA} -- An Authentic Dataset for Narrative Comprehension},\n publisher = {Association for Computational Linguistics},\n year = {2022}\n}",
13
  "contact-name": "Ying Xu, Dakuo Wang",
14
  "contact-email": "[email protected], [email protected]"
@@ -32,9 +32,9 @@
32
  "academic"
33
  ],
34
  "organization-names": "University of California Irvine",
35
- "creators": "Ying Xu(University of California Irvine); Dakuo Wang(IBM Research); Mo Yu(IBM Research); Daniel Ritchie(University of California Irvine); Bingsheng Yao(Rensselaer Polytechnic Institute); Tongshuang Wu(University of Washington); Zheng Zhang(University of Notre Dame); Toby Jia-Jun Li(University of Notre Dame); Nora Bradford(University of California Irvine); Branda Sun(University of California Irvine); Tran Bao Hoang(University of California Irvine); Yisi Sang(Syracuse University); Yufang Hou(IBM Research Ireland); Xiaojuan Ma(Hong Kong Univ. of Sci and Tech); Diyi Yang(Georgia Institute of Technology); Nanyun Peng(University of California Los Angeles); Zhou Yu(Columbia University); Mark Warschauer(University of California Irvine)",
36
  "funding": "Schmidt Futures",
37
- "gem-added-by": "Dakuo Wang(IBM Research); Bingsheng Yao(Rensselaer Polytechnic Institute); Ying Xu(University of California Irvine)"
38
  },
39
  "structure": {
40
  "data-fields": "- `story_name`: a string of the story name to which the story section content belongs. Full story data can be found [here](https://github.com/uci-soe/FairytaleQAData).\n\n- `content`: a string of the story section(s) content related to the experts' labeled QA-pair. Used as the input for both Question Generation and Question Answering tasks. \n\n- `question`: a string of the question content. Used as the input for Question Answering task and as the output for Question Generation task. \n\n- `answer`: a string of the answer content for all splits. Used as the input for Question Generation task and as the output for Question Answering task.\n\n- `gem_id`: a string of id follows GEM naming convention ```GEM-${DATASET_NAME}-${SPLIT-NAME}-${id}``` where id is an incrementing number starting at 1\n\n- `target`: a string of the question content being used for training\n\n- `references`: a list of string containing the question content being used for automatic eval\n\n- `local_or_sum`: a string of either local or summary, indicating whether the QA is related to one story section or multiple sections \n\n- `attribute`: a string of one of character, causal relationship, action, setting, feeling, prediction, or outcome resolution. Classification of the QA by education experts annotators via 7 narrative elements on an established framework\n \n- `ex_or_im`: a string of either explicit or implicit, indicating whether the answers can be directly found in the story content or cannot be directly from the story content.\n",
 
5
  },
6
  "where": {
7
  "has-leaderboard": "yes",
8
+ "leaderboard-url": "[PapersWithCode](https://paperswithcode.com/sota/question-generation-on-fairytaleqa)",
9
  "leaderboard-description": "The task was to generate questions corresponding to the given answers and the story context. Success on the Question Generation task is typically measured by achieving a high ROUGE-L score to the reference ground-truth question.",
10
+ "data-url": "[Github](https://github.com/uci-soe/FairytaleQAData)",
11
+ "paper-url": "[ArXiv](https://arxiv.org/abs/2203.13947)",
12
  "paper-bibtext": "@inproceedings{xu2022fairytaleqa,\n author={Xu, Ying and Wang, Dakuo and Yu, Mo and Ritchie, Daniel and Yao, Bingsheng and Wu, Tongshuang and Zhang, Zheng and Li, Toby Jia-Jun and Bradford, Nora and Sun, Branda and Hoang, Tran Bao and Sang, Yisi and Hou, Yufang and Ma, Xiaojuan and Yang, Diyi and Peng, Nanyun and Yu, Zhou and Warschauer, Mark},\n title = {Fantastic Questions and Where to Find Them: Fairytale{QA} -- An Authentic Dataset for Narrative Comprehension},\n publisher = {Association for Computational Linguistics},\n year = {2022}\n}",
13
  "contact-name": "Ying Xu, Dakuo Wang",
14
  "contact-email": "[email protected], [email protected]"
 
32
  "academic"
33
  ],
34
  "organization-names": "University of California Irvine",
35
+ "creators": "Ying Xu (University of California Irvine); Dakuo Wang (IBM Research); Mo Yu (IBM Research); Daniel Ritchie (University of California Irvine); Bingsheng Yao (Rensselaer Polytechnic Institute); Tongshuang Wu (University of Washington); Zheng Zhang (University of Notre Dame); Toby Jia-Jun Li (University of Notre Dame); Nora Bradford (University of California Irvine); Branda Sun (University of California Irvine); Tran Bao Hoang (University of California Irvine); Yisi Sang (Syracuse University); Yufang Hou (IBM Research Ireland); Xiaojuan Ma (Hong Kong Univ. of Sci and Tech); Diyi Yang (Georgia Institute of Technology); Nanyun Peng (University of California Los Angeles); Zhou Yu (Columbia University); Mark Warschauer (University of California Irvine)",
36
  "funding": "Schmidt Futures",
37
+ "gem-added-by": "Dakuo Wang (IBM Research); Bingsheng Yao (Rensselaer Polytechnic Institute); Ying Xu (University of California Irvine)"
38
  },
39
  "structure": {
40
  "data-fields": "- `story_name`: a string of the story name to which the story section content belongs. Full story data can be found [here](https://github.com/uci-soe/FairytaleQAData).\n\n- `content`: a string of the story section(s) content related to the experts' labeled QA-pair. Used as the input for both Question Generation and Question Answering tasks. \n\n- `question`: a string of the question content. Used as the input for Question Answering task and as the output for Question Generation task. \n\n- `answer`: a string of the answer content for all splits. Used as the input for Question Generation task and as the output for Question Answering task.\n\n- `gem_id`: a string of id follows GEM naming convention ```GEM-${DATASET_NAME}-${SPLIT-NAME}-${id}``` where id is an incrementing number starting at 1\n\n- `target`: a string of the question content being used for training\n\n- `references`: a list of string containing the question content being used for automatic eval\n\n- `local_or_sum`: a string of either local or summary, indicating whether the QA is related to one story section or multiple sections \n\n- `attribute`: a string of one of character, causal relationship, action, setting, feeling, prediction, or outcome resolution. Classification of the QA by education experts annotators via 7 narrative elements on an established framework\n \n- `ex_or_im`: a string of either explicit or implicit, indicating whether the answers can be directly found in the story content or cannot be directly from the story content.\n",
README.md CHANGED
@@ -1,96 +1,211 @@
1
  ---
2
  annotations_creators:
3
- - expert-generated
4
  language_creators:
5
- - found
6
  languages:
7
- - en
8
  licenses:
9
  - unknown
10
  multilinguality:
11
- - monolingual
12
  pretty_name: FairytaleQA
13
  size_categories:
14
- - 10K<n<100K
15
  source_datasets:
16
  - original
17
  task_categories:
18
  - question-generation
19
  task_ids:
20
- - abstractive-qg
21
  ---
22
 
23
- # Dataset Card for FairytaleQA
24
-
25
- ## Table of Contents
26
- - [Dataset Description](#dataset-description)
27
- - [Dataset Summary](#dataset-summary)
28
- - [Supported Tasks](#supported-tasks-and-leaderboards)
29
- - [Languages](#languages)
30
- - [Dataset Structure](#dataset-structure)
31
- - [Data Instances](#data-instances)
32
- - [Data Fields](#data-instances)
33
- - [Data Splits](#data-instances)
34
- - [Dataset Creation](#dataset-creation)
35
- - [Curation Rationale](#curation-rationale)
36
- - [Source Data](#source-data)
37
- - [Annotations](#annotations)
38
- - [Personal and Sensitive Information](#personal-and-sensitive-information)
39
- - [Considerations for Using the Data](#considerations-for-using-the-data)
40
- - [Social Impact of Dataset](#social-impact-of-dataset)
41
- - [Discussion of Biases](#discussion-of-biases)
42
- - [Other Known Limitations](#other-known-limitations)
43
- - [Additional Information](#additional-information)
44
- - [Dataset Curators](#dataset-curators)
45
- - [Licensing Information](#licensing-information)
46
- - [Citation Information](#citation-information)
47
 
48
  ## Dataset Description
49
 
50
  - **Homepage:** [Needs More Information]
51
  - **Repository:** https://github.com/uci-soe/FairytaleQAData
52
  - **Paper:** https://arxiv.org/abs/2203.13947
53
- - **Leaderboard:** https://paperswithcode.com/dataset/fairytaleqa
54
- - **Point of Contact:** [Ying Xu](mailto:[email protected])
55
 
56
- ### Dataset Summary
57
 
58
- The FairytaleQA Dataset is an English-language dataset focusing on narrative comprehension of kindergarten to eighth-grade students. Generated by educational experts based on an evidence-based theoretical framework, FairytaleQA consists of 10,580 explicit and implicit questions derived from 278 children-friendly stories, covering seven types of narrative elements or relations. This GEM version FairytaleQA Dataset specifically supports the Question Generation task.
59
 
60
- ### Supported Tasks and Leaderboards
61
 
62
- - `question-generation`: The dataset can be used to train a model for Question Generation. The task was to generate questions corresponding to the given answers and the story context. Success on this task is typically measured by achieving a high [ROUGE](https://huggingface.co/metrics/rouge) score to the reference ground-truth questions. A [BART-based model](https://huggingface.co/facebook/bart-large) currently achieves a [ROUGE-L of 0.527/0.527](https://github.com/uci-soe/FairytaleQAData) on valid/test splits. This task has an active leaderboard which can be found at <https://paperswithcode.com/sota/question-generation-on-fairytaleqa> and ranks models based on [ROUGE](https://huggingface.co/metrics/rouge) score.
63
 
 
 
 
 
 
 
64
 
65
- ### Languages
 
66
 
67
- The text in the dataset is in English. The associated BCP-47 code is `en`.
 
68
 
69
- ## Dataset Structure
70
 
71
- ### Data Instances
72
 
73
- A typical data point comprises a question, the corresponding story content, and one answer for train split or two answers for valid/test splits. Education expert annotators labeled whether the answer is locally relevant to one story section or requires summarization capabilities from multiple story sections and whether the answers are explicit (can be directly found in the stories) or implicit (cannot be directly found in the story text). Additionally, education expert annotators categorize the QA-pairs via 7 narrative elements from an established framework.
74
 
75
- For the original question file of each story, human annotator may provide more than one answer. However, to maintain evaluation fairness across all QAs, we only load the first answer here for the QG task.
 
 
76
 
77
- An example from the FairytaleQA test split looks as follows:
78
 
79
- ```
80
- {'story_name': 'self-did-it',
81
- 'content': '" what is your name ? " asked the girl from underground . " self is my name , " said the woman . that seemed a curious name to the girl , and she once more began to pull the fire apart . then the woman grew angry and began to scold , and built it all up again . thus they went on for a good while ; but at last , while they were in the midst of their pulling apart and building up of the fire , the woman upset the tar - barrel on the girl from underground . then the latter screamed and ran away , crying : " father , father ! self burned me ! " " nonsense , if self did it , then self must suffer for it ! " came the answer from below the hill .',
82
- 'answer': 'the woman told the girl her name was self .',
83
- 'question': "why did the girl's father think the girl burned herself ?",
84
- 'gem_id': 'GEM-FairytaleQA-test-1006',
85
- 'target': "why did the girl's father think the girl burned herself ?",
86
- 'references': ["why did the girl's father think the girl burned herself ?"],
87
- 'local_or_sum': 'local',
88
- 'attribute': 'causal relationship',
89
- 'ex_or_im': 'implicit'}
90
- ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
91
 
92
- ### Data Fields
93
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
94
  - `story_name`: a string of the story name to which the story section content belongs. Full story data can be found [here](https://github.com/uci-soe/FairytaleQAData).
95
 
96
  - `content`: a string of the story section(s) content related to the experts' labeled QA-pair. Used as the input for both Question Generation and Question Answering tasks.
@@ -112,94 +227,459 @@ An example from the FairytaleQA test split looks as follows:
112
  - `ex_or_im`: a string of either explicit or implicit, indicating whether the answers can be directly found in the story content or cannot be directly from the story content.
113
 
114
 
115
- ### Data Splits
116
 
117
- The data is split into a training, validation, and test split randomly but we control the ratio of QA-pair numbers in train:valid:test splits close to 8:1:1. The final split sizes are as follows:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
118
 
119
  | | Train | Validation | Test |
120
  | ----- | ----- | ----- | ----- |
121
  | # Books | 232 | 23 | 23 |
122
  | # QA-Pairs | 8548 | 1025 |1007 |
123
 
124
- ## Dataset Creation
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
125
 
126
- ### Curation Rationale
127
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
128
  FairytaleQA was built to focus on comprehension of narratives in the education domain, targeting students from kindergarten to eighth grade. We focus on narrative comprehension for 1. it is a high-level comprehension skill strongly predictive of reading achievement and plays a central role in daily life as people frequently encounter narratives in different forms, 2. narrative stories have a clear structure of specific elements and relations among these elements, and there are existing validated narrative comprehension frameworks around this structure, which provides a basis for developing the annotation schema for our dataset.
129
 
130
- <!--We employed education experts to generate and classify 10,580 QA-pairs based on a collection of 278 fairytale stories for young readers. -->
131
 
132
- ### Source Data
 
 
133
 
134
- #### Initial Data Collection and Normalization
135
 
136
- The narrative texts utilized in the dataset are classic fairytales with clear narrative structures. We gathered the text from the [Project Gutenberg](https://www.gutenberg.org/) website, using “fairytale” as the search term. Due to a large number of fairytales found, we used the most popular stories based on the number of downloads since these stories are presumably of higher quality.
 
 
137
 
138
- To ensure the readability of the text, we made a small number of minor revisions to some obviously outdated vocabulary (e.g., changing “ere” to “before”) and the unconventional use of punctuation (e.g., changing consecutive semi-colons to periods). For each story, we evaluated the reading difficulty level using the [textstat](https://pypi.org/project/textstat/) Python package, primarily based on sentence length, word length, and commonness of words. We excluded stories that are at 10th grade level or above.
139
 
140
- These texts were broken down into small sections based on their semantic content by our annotators. The annotators were instructed to split the story into sections of 100-300 words that also contain meaningful content and are separated at natural story breaks. An initial annotator would split the story, and this would be reviewed by a cross-checking annotator. Most of the resulting sections were one natural paragraph of the original text.
 
 
 
 
 
 
141
 
142
- #### Who are the source language producers?
143
 
 
 
 
 
 
 
 
 
144
  The fairytale story texts are from the [Project Gutenberg](https://www.gutenberg.org/) website
145
 
146
- ### Annotations
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
147
 
148
- #### Annotation process
149
 
 
 
150
  The annotators were instructed to imagine that they were creating questions to test elementary or middle school students in the process of reading a complete story. We required the annotators to generate only natural, open-ended questions, avoiding “yes-” or “no-” questions. We also instructed them to provide a diverse set of questions about 7 different narrative elements, and with both implicit and explicit questions.
151
 
152
  We asked the annotators to also generate answers for each of their questions. We asked them to provide the shortest possible answers but did not restrict them to complete sentences or short phrases. We also asked the annotators to label which section(s) the question and answer was from.
153
 
154
- All annotators received a two-week training in which each of them was familiarized with the coding template and conducted practice coding on the same five stories. The practice QA pairs were then reviewed by the other annotators and the three experts, and discrepancies among annotators were discussed. During the annotation process, the team met once every week to review and discuss each member’s work. All QA pairs were cross-checked by two annotators, and 10% of the QA pairs were additionally checked by the expert supervisor. For the 46 stories used as the evaluation set, we annotate a second reference answer by asking an annotator to independently read the story and answer the questions generated by others.
155
 
 
156
 
157
- #### Who are the annotators?
158
 
159
- Five annotators were involved in the annotation of QA pairs. All of these annotators have a B.A. degree in education, psychology, or cognitive science and have substantial experience in teaching and reading assessment. These annotators were supervised by three experts in literacy education.
160
 
161
- ### Personal and Sensitive Information
162
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
163
  [N/A]
164
 
165
- ## Considerations for Using the Data
166
 
167
- ### Social Impact of Dataset
 
 
 
 
 
 
 
168
 
169
- The purpose of this dataset is to help develop systems to facilitate assessment and training of narrative comprehension skills for children in education domain.
 
 
170
 
171
- The dataset distinguishes fine-grained reading skills, such as the understanding of varying narrative elements, and contains high-quality QA-pairs generated by education experts with sufficient training and education domain knowledge to create valid QA-pairs in a consistent way.
 
 
 
 
172
 
173
- This specific GEM version of our FairytaleQA dataset is for the Question Generation task because this dataset is suitable for developing models to automatically generate questions that satisfy the need for a continuous supply of new questions, which can potentially enable large-scale development of AI-supported interactive platforms for the learning and assessment of reading comprehension skills.
174
 
175
  ### Discussion of Biases
176
 
177
- This dataset could also be served to identify social stereotypes represented in story narratives. It will be valuable to analyze how social stereotypes are represented in the children’s literature collected in our dataset, thus enabling the development of automatic systems that detect and mitigate social biases. This type of bias analysis has been an underexplored research topic for the ML community, yet it will have profound societal impacts.
178
 
 
 
 
179
 
180
- ### Other Known Limitations
181
 
 
 
182
  [N/A]
183
 
184
- ## Additional Information
185
 
186
- ### Dataset Curators
187
 
188
- The dataset was created by Ying Xu, Dakuo Wang, Mo Yu, Daniel Ritchie, Bingsheng Yao, Tongshuang Wu, Zheng Zhang, Toby Jia-Jun Li, Nora Bradford, Branda Sun, Tran Bao Hoang, Yisi Sang, Yufang Hou, Xiaojuan Ma, Diyi Yang, Nanyun Peng, Zhou Yu, and Mark Warschauer.
189
 
190
- Schmidt Futures provided funding for the development of the FairytaleQA dataset.
191
 
192
- ### Licensing Information
193
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
194
  [N/A]
195
 
196
- ### Citation Information
197
 
198
- ```
199
- @inproceedings{xu2022fairytaleqa,
200
- author={Xu, Ying and Wang, Dakuo and Yu, Mo and Ritchie, Daniel and Yao, Bingsheng and Wu, Tongshuang and Zhang, Zheng and Li, Toby Jia-Jun and Bradford, Nora and Sun, Branda and Hoang, Tran Bao and Sang, Yisi and Hou, Yufang and Ma, Xiaojuan and Yang, Diyi and Peng, Nanyun and Yu, Zhou and Warschauer, Mark},
201
- title = {Fantastic Questions and Where to Find Them: Fairytale{QA} -- An Authentic Dataset for Narrative Comprehension},
202
- publisher = {Association for Computational Linguistics},
203
- year = {2022}
204
- }
205
- ```
 
1
  ---
2
  annotations_creators:
3
+ - expert-created
4
  language_creators:
5
+ - unknown
6
  languages:
7
+ - unknown
8
  licenses:
9
  - unknown
10
  multilinguality:
11
+ - unknown
12
  pretty_name: FairytaleQA
13
  size_categories:
14
+ - unknown
15
  source_datasets:
16
  - original
17
  task_categories:
18
  - question-generation
19
  task_ids:
20
+ - unknown
21
  ---
22
 
23
+ # Dataset Card for GEM/FairytaleQA
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
 
25
  ## Dataset Description
26
 
27
  - **Homepage:** [Needs More Information]
28
  - **Repository:** https://github.com/uci-soe/FairytaleQAData
29
  - **Paper:** https://arxiv.org/abs/2203.13947
30
+ - **Leaderboard:** https://paperswithcode.com/sota/question-generation-on-fairytaleqa
31
+ - **Point of Contact:** Ying Xu, Dakuo Wang
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/FairytaleQA).
36
 
37
+ ### Dataset Summary
38
 
39
+ The FairytaleQA Dataset is an English-language dataset focusing on narrative comprehension of kindergarten to eighth-grade students. Generated by educational experts based on an evidence-based theoretical framework, FairytaleQA consists of 10,580 explicit and implicit questions derived from 278 children-friendly stories, covering seven types of narrative elements or relations. The Dataset was corrected to support both the tasks of Question Generation and Question Answering.
40
 
41
+ You can load the dataset via:
42
+ ```
43
+ import datasets
44
+ data = datasets.load_dataset('GEM/FairytaleQA')
45
+ ```
46
+ The data loader can be found [here](https://huggingface.co/datasets/GEM/FairytaleQA).
47
 
48
+ #### paper
49
+ [ArXiv](https://arxiv.org/abs/2203.13947)
50
 
51
+ #### authors
52
+ Ying Xu (University of California Irvine); Dakuo Wang (IBM Research); Mo Yu (IBM Research); Daniel Ritchie (University of California Irvine); Bingsheng Yao (Rensselaer Polytechnic Institute); Tongshuang Wu (University of Washington); Zheng Zhang (University of Notre Dame); Toby Jia-Jun Li (University of Notre Dame); Nora Bradford (University of California Irvine); Branda Sun (University of California Irvine); Tran Bao Hoang (University of California Irvine); Yisi Sang (Syracuse University); Yufang Hou (IBM Research Ireland); Xiaojuan Ma (Hong Kong Univ. of Sci and Tech); Diyi Yang (Georgia Institute of Technology); Nanyun Peng (University of California Los Angeles); Zhou Yu (Columbia University); Mark Warschauer (University of California Irvine)
53
 
54
+ ## Dataset Overview
55
 
56
+ ### Where to find the Data and its Documentation
57
 
58
+ #### Download
59
 
60
+ <!-- info: What is the link to where the original dataset is hosted? -->
61
+ <!-- scope: telescope -->
62
+ [Github](https://github.com/uci-soe/FairytaleQAData)
63
 
64
+ #### Paper
65
 
66
+ <!-- info: What is the link to the paper describing the dataset (open access preferred)? -->
67
+ <!-- scope: telescope -->
68
+ [ArXiv](https://arxiv.org/abs/2203.13947)
69
+
70
+ #### BibTex
71
+
72
+ <!-- info: Provide the BibTex-formatted reference for the dataset. Please use the correct published version (ACL anthology, etc.) instead of google scholar created Bibtex. -->
73
+ <!-- scope: microscope -->
74
+ @inproceedings{xu2022fairytaleqa,
75
+ author={Xu, Ying and Wang, Dakuo and Yu, Mo and Ritchie, Daniel and Yao, Bingsheng and Wu, Tongshuang and Zhang, Zheng and Li, Toby Jia-Jun and Bradford, Nora and Sun, Branda and Hoang, Tran Bao and Sang, Yisi and Hou, Yufang and Ma, Xiaojuan and Yang, Diyi and Peng, Nanyun and Yu, Zhou and Warschauer, Mark},
76
+ title = {Fantastic Questions and Where to Find Them: Fairytale{QA} -- An Authentic Dataset for Narrative Comprehension},
77
+ publisher = {Association for Computational Linguistics},
78
+ year = {2022}
79
+ }
80
+
81
+ #### Contact Name
82
+
83
+ <!-- quick -->
84
+ <!-- info: If known, provide the name of at least one person the reader can contact for questions about the dataset. -->
85
+ <!-- scope: periscope -->
86
+ Ying Xu, Dakuo Wang
87
+
88
+ #### Contact Email
89
+
90
+ <!-- info: If known, provide the email of at least one person the reader can contact for questions about the dataset. -->
91
+ <!-- scope: periscope -->
92
93
+
94
+ #### Has a Leaderboard?
95
+
96
+ <!-- info: Does the dataset have an active leaderboard? -->
97
+ <!-- scope: telescope -->
98
+ yes
99
+
100
+ #### Leaderboard Link
101
+
102
+ <!-- info: Provide a link to the leaderboard. -->
103
+ <!-- scope: periscope -->
104
+ [PapersWithCode](https://paperswithcode.com/sota/question-generation-on-fairytaleqa)
105
+
106
+ #### Leaderboard Details
107
+
108
+ <!-- info: Briefly describe how the leaderboard evaluates models. -->
109
+ <!-- scope: microscope -->
110
+ The task was to generate questions corresponding to the given answers and the story context. Success on the Question Generation task is typically measured by achieving a high ROUGE-L score to the reference ground-truth question.
111
+
112
+
113
+ ### Languages and Intended Use
114
+
115
+ #### Multilingual?
116
+
117
+ <!-- quick -->
118
+ <!-- info: Is the dataset multilingual? -->
119
+ <!-- scope: telescope -->
120
+ no
121
+
122
+ #### Covered Dialects
123
+
124
+ <!-- info: What dialects are covered? Are there multiple dialects per language? -->
125
+ <!-- scope: periscope -->
126
+ [N/A]
127
+
128
+ #### Covered Languages
129
+
130
+ <!-- quick -->
131
+ <!-- info: What languages/dialects are covered in the dataset? -->
132
+ <!-- scope: telescope -->
133
+ `English`
134
+
135
+ #### Whose Language?
136
+
137
+ <!-- info: Whose language is in the dataset? -->
138
+ <!-- scope: periscope -->
139
+ [N/A]
140
+
141
+ #### License
142
+
143
+ <!-- quick -->
144
+ <!-- info: What is the license of the dataset? -->
145
+ <!-- scope: telescope -->
146
+ unknown: License information unavailable
147
+
148
+ #### Intended Use
149
+
150
+ <!-- info: What is the intended use of the dataset? -->
151
+ <!-- scope: microscope -->
152
+ The purpose of this dataset is to help develop systems to facilitate assessment and training of narrative comprehension skills for children in education domain. The dataset distinguishes fine-grained reading skills, such as the understanding of varying narrative elements, and contains high-quality QA-pairs generated by education experts with sufficient training and education domain knowledge to create valid QA-pairs in a consistent way.
153
+
154
+ This dataset is suitable for developing models to automatically generate questions and QA-Pairs that satisfy the need for a continuous supply of new questions, which can potentially enable large-scale development of AI-supported interactive platforms for the learning and assessment of reading comprehension skills.
155
+
156
+ #### Primary Task
157
+
158
+ <!-- info: What primary task does the dataset support? -->
159
+ <!-- scope: telescope -->
160
+ Question Generation
161
+
162
+ #### Communicative Goal
163
+
164
+ <!-- quick -->
165
+ <!-- info: Provide a short description of the communicative goal of a model trained for this task on this dataset. -->
166
+ <!-- scope: periscope -->
167
+ The task was to generate questions corresponding to the given answers and the story context. Models trained for this task can potentially enable large-scale development of AI-supported interactive platforms for the learning and assessment of reading comprehension skills.
168
 
 
169
 
170
+ ### Credit
171
+
172
+ #### Curation Organization Type(s)
173
+
174
+ <!-- info: In what kind of organization did the dataset curation happen? -->
175
+ <!-- scope: telescope -->
176
+ `academic`
177
+
178
+ #### Curation Organization(s)
179
+
180
+ <!-- info: Name the organization(s). -->
181
+ <!-- scope: periscope -->
182
+ University of California Irvine
183
+
184
+ #### Dataset Creators
185
+
186
+ <!-- info: Who created the original dataset? List the people involved in collecting the dataset and their affiliation(s). -->
187
+ <!-- scope: microscope -->
188
+ Ying Xu (University of California Irvine); Dakuo Wang (IBM Research); Mo Yu (IBM Research); Daniel Ritchie (University of California Irvine); Bingsheng Yao (Rensselaer Polytechnic Institute); Tongshuang Wu (University of Washington); Zheng Zhang (University of Notre Dame); Toby Jia-Jun Li (University of Notre Dame); Nora Bradford (University of California Irvine); Branda Sun (University of California Irvine); Tran Bao Hoang (University of California Irvine); Yisi Sang (Syracuse University); Yufang Hou (IBM Research Ireland); Xiaojuan Ma (Hong Kong Univ. of Sci and Tech); Diyi Yang (Georgia Institute of Technology); Nanyun Peng (University of California Los Angeles); Zhou Yu (Columbia University); Mark Warschauer (University of California Irvine)
189
+
190
+ #### Funding
191
+
192
+ <!-- info: Who funded the data creation? -->
193
+ <!-- scope: microscope -->
194
+ Schmidt Futures
195
+
196
+ #### Who added the Dataset to GEM?
197
+
198
+ <!-- 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. -->
199
+ <!-- scope: microscope -->
200
+ Dakuo Wang (IBM Research); Bingsheng Yao (Rensselaer Polytechnic Institute); Ying Xu (University of California Irvine)
201
+
202
+
203
+ ### Dataset Structure
204
+
205
+ #### Data Fields
206
+
207
+ <!-- info: List and describe the fields present in the dataset. -->
208
+ <!-- scope: telescope -->
209
  - `story_name`: a string of the story name to which the story section content belongs. Full story data can be found [here](https://github.com/uci-soe/FairytaleQAData).
210
 
211
  - `content`: a string of the story section(s) content related to the experts' labeled QA-pair. Used as the input for both Question Generation and Question Answering tasks.
 
227
  - `ex_or_im`: a string of either explicit or implicit, indicating whether the answers can be directly found in the story content or cannot be directly from the story content.
228
 
229
 
230
+ #### Reason for Structure
231
 
232
+ <!-- info: How was the dataset structure determined? -->
233
+ <!-- scope: microscope -->
234
+ [N/A]
235
+
236
+ #### How were labels chosen?
237
+
238
+ <!-- info: How were the labels chosen? -->
239
+ <!-- scope: microscope -->
240
+ A typical data point comprises a question, the corresponding story content, and one answer. Education expert annotators labeled whether the answer is locally relevant to one story section or requires summarization capabilities from multiple story sections, and whether the answers are explicit (can be directly found in the stories) or implicit (cannot be directly found in the story text). Additionally, education expert annotators categorize the QA-pairs via 7 narrative elements from an establish framework.
241
+
242
+ #### Example Instance
243
+
244
+ <!-- info: Provide a JSON formatted example of a typical instance in the dataset. -->
245
+ <!-- scope: periscope -->
246
+ {'story_name': 'self-did-it',
247
+ 'content': '" what is your name ? " asked the girl from underground . " self is my name , " said the woman . that seemed a curious name to the girl , and she once more began to pull the fire apart . then the woman grew angry and began to scold , and built it all up again . thus they went on for a good while ; but at last , while they were in the midst of their pulling apart and building up of the fire , the woman upset the tar - barrel on the girl from underground . then the latter screamed and ran away , crying : " father , father ! self burned me ! " " nonsense , if self did it , then self must suffer for it ! " came the answer from below the hill .',
248
+ 'answer': 'the woman told the girl her name was self .',
249
+ 'question': "why did the girl's father think the girl burned herself ?",
250
+ 'gem_id': 'GEM-FairytaleQA-test-1006',
251
+ 'target': "why did the girl's father think the girl burned herself ?",
252
+ 'references': ["why did the girl's father think the girl burned herself ?"],
253
+ 'local_or_sum': 'local',
254
+ 'attribute': 'causal relationship',
255
+ 'ex_or_im': 'implicit'}
256
+
257
+ #### Data Splits
258
+
259
+ <!-- info: Describe and name the splits in the dataset if there are more than one. -->
260
+ <!-- scope: periscope -->
261
+ The data is split into a train, validation, and test split randomly. The final split sizes are as follows:
262
 
263
  | | Train | Validation | Test |
264
  | ----- | ----- | ----- | ----- |
265
  | # Books | 232 | 23 | 23 |
266
  | # QA-Pairs | 8548 | 1025 |1007 |
267
 
268
+ #### Splitting Criteria
269
+
270
+ <!-- info: Describe any criteria for splitting the data, if used. If there are differences between the splits (e.g., if the training annotations are machine-generated and the dev and test ones are created by humans, or if different numbers of annotators contributed to each example), describe them here. -->
271
+ <!-- scope: microscope -->
272
+ The books are randomly split into train/validation/test splits. We control the ratio of QA-pair numbers in train:validation:test splits close to 8:1:1
273
+
274
+ ####
275
+
276
+ <!-- info: What does an outlier of the dataset in terms of length/perplexity/embedding look like? -->
277
+ <!-- scope: microscope -->
278
+ [N/A]
279
+
280
+
281
+
282
+ ## Dataset in GEM
283
+
284
+ ### Rationale for Inclusion in GEM
285
+
286
+ #### Why is the Dataset in GEM?
287
+
288
+ <!-- info: What does this dataset contribute toward better generation evaluation and why is it part of GEM? -->
289
+ <!-- scope: microscope -->
290
+ The dataset distinguishes fine-grained reading skills, such as the understanding of varying narrative elements, and contains high-quality QA-pairs generated by education experts with sufficient training and education domain knowledge to create valid QA-pairs in a consistent way.
291
+
292
+
293
+
294
+ #### Similar Datasets
295
+
296
+ <!-- info: Do other datasets for the high level task exist? -->
297
+ <!-- scope: telescope -->
298
+ no
299
+
300
+ #### Ability that the Dataset measures
301
+
302
+ <!-- info: What aspect of model ability can be measured with this dataset? -->
303
+ <!-- scope: periscope -->
304
+ This dataset is suitable for developing models to automatically generate questions or QA-pairs that satisfy the need for a continuous supply of new questions, which can potentially enable large-scale development of AI-supported interactive platforms for the learning and assessment of reading comprehension skills.
305
+
306
+
307
+ ### GEM-Specific Curation
308
+
309
+ #### Modificatied for GEM?
310
+
311
+ <!-- info: Has the GEM version of the dataset been modified in any way (data, processing, splits) from the original curated data? -->
312
+ <!-- scope: telescope -->
313
+ yes
314
+
315
+ #### GEM Modifications
316
+
317
+ <!-- info: What changes have been made to he original dataset? -->
318
+ <!-- scope: periscope -->
319
+ `data points removed`
320
+
321
+ #### Modification Details
322
+
323
+ <!-- info: For each of these changes, described them in more details and provided the intended purpose of the modification -->
324
+ <!-- scope: microscope -->
325
+ The original data contains two answers by different annotators in validation/test splits, we removed the 2nd answer for GEM version because it is not being used for the Question Generation task.
326
+
327
+ #### Additional Splits?
328
+
329
+ <!-- info: Does GEM provide additional splits to the dataset? -->
330
+ <!-- scope: telescope -->
331
+ no
332
 
 
333
 
334
+ ### Getting Started with the Task
335
+
336
+ #### Pointers to Resources
337
+
338
+ <!-- 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. -->
339
+ <!-- scope: microscope -->
340
+ [N/A]
341
+
342
+
343
+
344
+ ## Previous Results
345
+
346
+ ### Previous Results
347
+
348
+ #### Measured Model Abilities
349
+
350
+ <!-- info: What aspect of model ability can be measured with this dataset? -->
351
+ <!-- scope: telescope -->
352
+ We are able to measure model's capabilities of generating various types of questions that corresponds to different narrative elements with the FairytaleQA dataset on the Question Generation Task
353
+
354
+ #### Metrics
355
+
356
+ <!-- info: What metrics are typically used for this task? -->
357
+ <!-- scope: periscope -->
358
+ `ROUGE`
359
+
360
+ #### Proposed Evaluation
361
+
362
+ <!-- info: List and describe the purpose of the metrics and evaluation methodology (including human evaluation) that the dataset creators used when introducing this task. -->
363
+ <!-- scope: microscope -->
364
+ The task was to generate questions corresponding to the given answers and the story context. Success on this task is typically measured by achieving a high [ROUGE](https://huggingface.co/metrics/rouge) score to the reference ground-truth questions.
365
+
366
+ #### Previous results available?
367
+
368
+ <!-- info: Are previous results available? -->
369
+ <!-- scope: telescope -->
370
+ yes
371
+
372
+ #### Relevant Previous Results
373
+
374
+ <!-- info: What are the most relevant previous results for this task/dataset? -->
375
+ <!-- scope: microscope -->
376
+ A [BART-based model](https://huggingface.co/facebook/bart-large) currently achieves a [ROUGE-L of 0.527/0.527](https://github.com/uci-soe/FairytaleQAData) on valid/test splits, which is reported as the baseline experiment for the dataset [paper](https://arxiv.org/pdf/2203.13947.pdf).
377
+
378
+
379
+
380
+ ## Dataset Curation
381
+
382
+ ### Original Curation
383
+
384
+ #### Original Curation Rationale
385
+
386
+ <!-- info: Original curation rationale -->
387
+ <!-- scope: telescope -->
388
  FairytaleQA was built to focus on comprehension of narratives in the education domain, targeting students from kindergarten to eighth grade. We focus on narrative comprehension for 1. it is a high-level comprehension skill strongly predictive of reading achievement and plays a central role in daily life as people frequently encounter narratives in different forms, 2. narrative stories have a clear structure of specific elements and relations among these elements, and there are existing validated narrative comprehension frameworks around this structure, which provides a basis for developing the annotation schema for our dataset.
389
 
390
+ #### Communicative Goal
391
 
392
+ <!-- info: What was the communicative goal? -->
393
+ <!-- scope: periscope -->
394
+ The purpose of this dataset is to help develop systems to facilitate assessment and training of narrative comprehension skills for children in education domain.
395
 
396
+ #### Sourced from Different Sources
397
 
398
+ <!-- info: Is the dataset aggregated from different data sources? -->
399
+ <!-- scope: telescope -->
400
+ no
401
 
 
402
 
403
+ ### Language Data
404
+
405
+ #### How was Language Data Obtained?
406
+
407
+ <!-- info: How was the language data obtained? -->
408
+ <!-- scope: telescope -->
409
+ `Found`
410
 
411
+ #### Where was it found?
412
 
413
+ <!-- info: If found, where from? -->
414
+ <!-- scope: telescope -->
415
+ `Single website`
416
+
417
+ #### Language Producers
418
+
419
+ <!-- info: What further information do we have on the language producers? -->
420
+ <!-- scope: microscope -->
421
  The fairytale story texts are from the [Project Gutenberg](https://www.gutenberg.org/) website
422
 
423
+ #### Topics Covered
424
+
425
+ <!-- info: Does the language in the dataset focus on specific topics? How would you describe them? -->
426
+ <!-- scope: periscope -->
427
+ We gathered the text from the Project Gutenberg website, using “fairytale” as the search term.
428
+
429
+ #### Data Validation
430
+
431
+ <!-- info: Was the text validated by a different worker or a data curator? -->
432
+ <!-- scope: telescope -->
433
+ validated by data curator
434
+
435
+ #### Data Preprocessing
436
+
437
+ <!-- info: How was the text data pre-processed? (Enter N/A if the text was not pre-processed) -->
438
+ <!-- scope: microscope -->
439
+ Due to a large number of fairytales found, we used the most popular stories based on the number of downloads since these stories are presumably of higher quality. To ensure the readability of the text, we made a small number of minor revisions to some obviously outdated vocabulary (e.g., changing “ere” to “before”) and the unconventional use of punctuation (e.g., changing consecutive semi-colons to periods).
440
+
441
+ These texts were broken down into small sections based on their semantic content by our annotators. The annotators were instructed to split the story into sections of 100-300 words that also contain meaningful content and are separated at natural story breaks. An initial annotator would split the story, and this would be reviewed by a cross-checking annotator. Most of the resulting sections were one natural paragraph of the original text.
442
+
443
+ #### Was Data Filtered?
444
+
445
+ <!-- info: Were text instances selected or filtered? -->
446
+ <!-- scope: telescope -->
447
+ manually
448
+
449
+ #### Filter Criteria
450
+
451
+ <!-- info: What were the selection criteria? -->
452
+ <!-- scope: microscope -->
453
+ For each story, we evaluated the reading difficulty level using the [textstat](https://pypi.org/project/textstat/) Python package, primarily based on sentence length, word length, and commonness of words. We excluded stories that are at 10th grade level or above.
454
+
455
+
456
+ ### Structured Annotations
457
+
458
+ #### Additional Annotations?
459
+
460
+ <!-- quick -->
461
+ <!-- info: Does the dataset have additional annotations for each instance? -->
462
+ <!-- scope: telescope -->
463
+ expert created
464
+
465
+ #### Number of Raters
466
+
467
+ <!-- info: What is the number of raters -->
468
+ <!-- scope: telescope -->
469
+ 2<n<10
470
+
471
+ #### Rater Qualifications
472
+
473
+ <!-- info: Describe the qualifications required of an annotator. -->
474
+ <!-- scope: periscope -->
475
+ All of these annotators have a B.A. degree in education, psychology, or cognitive science and have substantial experience in teaching and reading assessment. These annotators were supervised by three experts in literacy education.
476
+
477
+ #### Raters per Training Example
478
+
479
+ <!-- info: How many annotators saw each training example? -->
480
+ <!-- scope: periscope -->
481
+ 2
482
+
483
+ #### Raters per Test Example
484
+
485
+ <!-- info: How many annotators saw each test example? -->
486
+ <!-- scope: periscope -->
487
+ 3
488
+
489
+ #### Annotation Service?
490
+
491
+ <!-- info: Was an annotation service used? -->
492
+ <!-- scope: telescope -->
493
+ no
494
+
495
+ #### Annotation Values
496
+
497
+ <!-- info: Purpose and values for each annotation -->
498
+ <!-- scope: microscope -->
499
+ The dataset annotation distinguishes fine-grained reading skills, such as the understanding of varying narrative elements, and contains high-quality QA-pairs generated by education experts with sufficient training and education domain knowledge to create valid QA-pairs in a consistent way.
500
+
501
+ #### Any Quality Control?
502
+
503
+ <!-- info: Quality control measures? -->
504
+ <!-- scope: telescope -->
505
+ validated by data curators
506
 
507
+ #### Quality Control Details
508
 
509
+ <!-- info: Describe the quality control measures that were taken. -->
510
+ <!-- scope: microscope -->
511
  The annotators were instructed to imagine that they were creating questions to test elementary or middle school students in the process of reading a complete story. We required the annotators to generate only natural, open-ended questions, avoiding “yes-” or “no-” questions. We also instructed them to provide a diverse set of questions about 7 different narrative elements, and with both implicit and explicit questions.
512
 
513
  We asked the annotators to also generate answers for each of their questions. We asked them to provide the shortest possible answers but did not restrict them to complete sentences or short phrases. We also asked the annotators to label which section(s) the question and answer was from.
514
 
515
+ All annotators received a two-week training in which each of them was familiarized with the coding template and conducted practice coding on the same five stories. The practice QA pairs were then reviewed by the other annotators and the three experts, and discrepancies among annotators were discussed. During the annotation process, the team met once every week to review and discuss each member’s work. All QA pairs were cross-checked by two annotators, and 10% of the QA pairs were additionally checked by the expert supervisor.
516
 
517
+ For the 46 stories used as the evaluation set, we annotate a second reference answer by asking an annotator to independently read the story and answer the questions generated by others.
518
 
 
519
 
520
+ ### Consent
521
 
522
+ #### Any Consent Policy?
523
 
524
+ <!-- info: Was there a consent policy involved when gathering the data? -->
525
+ <!-- scope: telescope -->
526
+ yes
527
+
528
+ #### Consent Policy Details
529
+
530
+ <!-- info: What was the consent policy? -->
531
+ <!-- scope: microscope -->
532
+ During the annotation process, the team met once every week to review and discuss each member’s work. All QA pairs were cross-checked by two annotators, and 10% of the QA pairs were additionally checked by the expert supervisor.
533
+
534
+ #### Other Consented Downstream Use
535
+
536
+ <!-- info: What other downstream uses of the data did the original data creators and the data curators consent to? -->
537
+ <!-- scope: microscope -->
538
+ Aside from Question Generation task, the data creators and curators used this data for Question Answering, and QA-Pair Generation tasks, and to identify social stereotypes represented in story narratives.
539
+
540
+
541
+ ### Private Identifying Information (PII)
542
+
543
+ #### Contains PII?
544
+
545
+ <!-- quick -->
546
+ <!-- info: Does the source language data likely contain Personal Identifying Information about the data creators or subjects? -->
547
+ <!-- scope: telescope -->
548
+ no PII
549
+
550
+ #### Justification for no PII
551
+
552
+ <!-- info: Provide a justification for selecting `no PII` above. -->
553
+ <!-- scope: periscope -->
554
+ The story content is from publically available knowledge website and the annotated QA-pairs are about general knowledge to the story content without references to the author or to any persons
555
+
556
+
557
+ ### Maintenance
558
+
559
+ #### Any Maintenance Plan?
560
+
561
+ <!-- info: Does the original dataset have a maintenance plan? -->
562
+ <!-- scope: telescope -->
563
+ yes
564
+
565
+ #### Maintenance Plan Details
566
+
567
+ <!-- info: Describe the original dataset's maintenance plan. -->
568
+ <!-- scope: microscope -->
569
+ We plan to host various splits for the FairytaleQA dataset to better serve various types of research interests. We have the original data for 2 different split approaches including train/validation/test splits and split by fairytale origins. We are also plan to host the dataset on multiple platforms for various tasks.
570
+
571
+ #### Maintainer Contact Information
572
+
573
+ <!-- info: Provide contact information of a person responsible for the dataset maintenance -->
574
+ <!-- scope: periscope -->
575
+ Daniel Ritchie
576
+
577
+ #### Any Contestation Mechanism?
578
+
579
+ <!-- info: Does the maintenance plan include a contestation mechanism allowing individuals to request removal fo content? -->
580
+ <!-- scope: periscope -->
581
+ no mechanism
582
+
583
+
584
+
585
+ ## Broader Social Context
586
+
587
+ ### Previous Work on the Social Impact of the Dataset
588
+
589
+ #### Usage of Models based on the Data
590
+
591
+ <!-- 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? -->
592
+ <!-- scope: telescope -->
593
+ yes - models trained on this dataset
594
+
595
+ #### Social Impact Observations
596
+
597
+ <!-- info: Did any of these previous uses result in observations about the social impact of the systems? In particular, has there been work outlining the risks and limitations of the system? Provide links and descriptions here. -->
598
+ <!-- scope: microscope -->
599
  [N/A]
600
 
601
+ #### Changes as Consequence of Social Impact
602
 
603
+ <!-- info: Have any changes been made to the dataset as a result of these observations? -->
604
+ <!-- scope: periscope -->
605
+ [N/A]
606
+
607
+
608
+ ### Impact on Under-Served Communities
609
+
610
+ #### Addresses needs of underserved Communities?
611
 
612
+ <!-- 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). -->
613
+ <!-- scope: telescope -->
614
+ yes
615
 
616
+ #### Details on how Dataset Addresses the Needs
617
+
618
+ <!-- info: Describe how this dataset addresses the needs of underserved communities. -->
619
+ <!-- scope: microscope -->
620
+ From the educational perspective, given that reading comprehension is a multicomponent skill, it is ideal for comprehension questions to be able to identify students’ performance in specific sub-skills, thus allowing teachers to provide tailored guidance.
621
 
 
622
 
623
  ### Discussion of Biases
624
 
625
+ #### Any Documented Social Biases?
626
 
627
+ <!-- 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. -->
628
+ <!-- scope: telescope -->
629
+ unsure
630
 
631
+ #### Are the Language Producers Representative of the Language?
632
 
633
+ <!-- info: Does the distribution of language producers in the dataset accurately represent the full distribution of speakers of the language world-wide? If not, how does it differ? -->
634
+ <!-- scope: periscope -->
635
  [N/A]
636
 
 
637
 
 
638
 
639
+ ## Considerations for Using the Data
640
 
641
+ ### PII Risks and Liability
642
 
643
+ #### Potential PII Risk
644
 
645
+ <!-- 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. -->
646
+ <!-- scope: microscope -->
647
+ [N/A]
648
+
649
+
650
+ ### Licenses
651
+
652
+ #### Copyright Restrictions on the Dataset
653
+
654
+ <!-- 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? -->
655
+ <!-- scope: periscope -->
656
+ `research use only`
657
+
658
+ #### Copyright Restrictions on the Language Data
659
+
660
+ <!-- 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? -->
661
+ <!-- scope: periscope -->
662
+ `public domain`
663
+
664
+
665
+ ### Known Technical Limitations
666
+
667
+ #### Technical Limitations
668
+
669
+ <!-- info: Describe any known technical limitations, such as spurrious correlations, train/test overlap, annotation biases, or mis-annotations, and cite the works that first identified these limitations when possible. -->
670
+ <!-- scope: microscope -->
671
+ We noticed that human results are obtained via cross-estimation between the two annotated answers, thus are underestimated. One possibility for future work is to conduct a large-scale human annotation to collect more answers per question and then leverage the massively annotated answers to better establish a human performance evaluation.
672
+
673
+ #### Unsuited Applications
674
+
675
+ <!-- info: When using a model trained on this dataset in a setting where users or the public may interact with its predictions, what are some pitfalls to look out for? In particular, describe some applications of the general task featured in this dataset that its curation or properties make it less suitable for. -->
676
+ <!-- scope: microscope -->
677
+ The QA-pairs annotated by education experts are targeting the audience of children from kindergarten to eighth grade, so the difficulty of QA-pairs are not suitable to compare with other existing dataset that are sourced from knowledge graphs or knowledge bases like Wikipedia.
678
+
679
+ #### Discouraged Use Cases
680
+
681
+ <!-- info: What are some discouraged use cases of a model trained to maximize the proposed metrics on this dataset? In particular, think about settings where decisions made by a model that performs reasonably well on the metric my still have strong negative consequences for user or members of the public. -->
682
+ <!-- scope: microscope -->
683
  [N/A]
684
 
 
685