gem

  • Description:

GEM is a benchmark environment for Natural Language Generation with a focus on its Evaluation, both through human annotations and automated Metrics.

GEM aims to: (1) measure NLG progress across 13 datasets spanning many NLG tasks and languages. (2) provide an in-depth analysis of data and models presented via data statements and challenge sets. (3) develop standards for evaluation of generated text using both automated and human metrics.

More information can be found at https://gem-benchmark.com.

gem/common_gen (default config)

  • Config description: CommonGen is a constrained text generation task, associated with a benchmark dataset, to explicitly test machines for the ability of generative commonsense reasoning. Given a set of common concepts; the task is to generate a coherent sentence describing an everyday scenario using these concepts.

  • Download size: 1.84 MiB

  • Dataset size: 16.84 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 1,497
'train' 67,389
'validation' 993
  • Feature structure:
FeaturesDict({
    'concept_set_id': int32,
    'concepts': Sequence(string),
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'target': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
concept_set_id Tensor int32
concepts Sequence(Tensor) (None,) string
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
target Tensor string
  • Citation:
@inproceedings{lin2020commongen,
  title = "CommonGen: A Constrained Text Generation Challenge for Generative Commonsense Reasoning",
  author = "Lin, Bill Yuchen  and
    Zhou, Wangchunshu  and
    Shen, Ming  and
    Zhou, Pei  and
    Bhagavatula, Chandra  and
    Choi, Yejin  and
    Ren, Xiang",
  booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2020",
  month = nov,
  year = "2020",
  address = "Online",
  publisher = "Association for Computational Linguistics",
  url = "https://www.aclweb.org/anthology/2020.findings-emnlp.165",
  pages = "1823--1840",
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/cs_restaurants

  • Config description: The task is generating responses in the context of a (hypothetical) dialogue system that provides information about restaurants. The input is a basic intent/dialogue act type and a list of slots (attributes) and their values. The output is a natural language sentence.

  • Download size: 1.46 MiB

  • Dataset size: 2.71 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 842
'train' 3,569
'validation' 781
  • Feature structure:
FeaturesDict({
    'dialog_act': string,
    'dialog_act_delexicalized': string,
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'target': string,
    'target_delexicalized': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
dialog_act Tensor string
dialog_act_delexicalized Tensor string
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
target Tensor string
target_delexicalized Tensor string
  • Citation:
@inproceedings{cs_restaurants,
  address = {Tokyo, Japan},
  title = {Neural {Generation} for {Czech}: {Data} and {Baselines} },
  shorttitle = {Neural {Generation} for {Czech} },
  url = {https://www.aclweb.org/anthology/W19-8670/},
  urldate = {2019-10-18},
  booktitle = {Proceedings of the 12th {International} {Conference} on {Natural} {Language} {Generation} ({INLG} 2019)},
  author = {Dušek, Ondřej and Jurčíček, Filip},
  month = oct,
  year = {2019},
  pages = {563--574}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/dart

  • Config description: DART is a large and open-domain structured DAta Record to Text generation corpus with high-quality sentence annotations with each input being a set of entity-relation triples following a tree-structured ontology.

  • Download size: 28.01 MiB

  • Dataset size: 33.78 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'test' 6,959
'train' 62,659
'validation' 2,768
  • Feature structure:
FeaturesDict({
    'dart_id': int32,
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'subtree_was_extended': bool,
    'target': string,
    'target_sources': Sequence(string),
    'tripleset': Sequence(string),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
dart_id Tensor int32
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
subtree_was_extended Tensor bool
target Tensor string
target_sources Sequence(Tensor) (None,) string
tripleset Sequence(Tensor) (None,) string
  • Citation:
@article{radev2020dart,
  title=Dart: Open-domain structured data record to text generation,
  author={Radev, Dragomir and Zhang, Rui and Rau, Amrit and Sivaprasad, Abhinand and Hsieh, Chiachun and Rajani, Nazneen Fatema and Tang, Xiangru and Vyas, Aadit and Verma, Neha and Krishna, Pranav and others},
  journal={arXiv preprint arXiv:2007.02871},
  year={2020}
}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/e2e_nlg

  • Config description: The E2E dataset is designed for a limited-domain data-to-text task -- generation of restaurant descriptions/recommendations based on up to 8 different attributes (name, area, price range etc.)

  • Download size: 13.99 MiB

  • Dataset size: 16.92 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 4,693
'train' 33,525
'validation' 4,299
  • Feature structure:
FeaturesDict({
    'gem_id': string,
    'gem_parent_id': string,
    'meaning_representation': string,
    'references': Sequence(string),
    'target': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
gem_id Tensor string
gem_parent_id Tensor string
meaning_representation Tensor string
references Sequence(Tensor) (None,) string
target Tensor string
  • Citation:
@inproceedings{e2e_cleaned,
  address = {Tokyo, Japan},
  title = {Semantic {Noise} {Matters} for {Neural} {Natural} {Language} {Generation} },
  url = {https://www.aclweb.org/anthology/W19-8652/},
  booktitle = {Proceedings of the 12th {International} {Conference} on {Natural} {Language} {Generation} ({INLG} 2019)},
  author = {Dušek, Ondřej and Howcroft, David M and Rieser, Verena},
  year = {2019},
  pages = {421--426},
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/mlsum_de

  • Config description: MLSum is a large-scale multiLingual summarization dataset. It is buillt from online news outlets, this split focusing on German.

  • Download size: 345.98 MiB

  • Dataset size: 963.60 MiB

  • Auto-cached (documentation): No

  • Splits:

Split Examples
'challenge_test_covid' 5,058
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 10,695
'train' 220,748
'validation' 11,392
  • Feature structure:
FeaturesDict({
    'date': string,
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'target': string,
    'text': string,
    'title': string,
    'topic': string,
    'url': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
date Tensor string
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
target Tensor string
text Tensor string
title Tensor string
topic Tensor string
url Tensor string
  • Citation:
@inproceedings{scialom-etal-2020-mlsum,
    title = "{MLSUM}: The Multilingual Summarization Corpus",
    author = {Scialom, Thomas  and Dray, Paul-Alexis  and Lamprier, Sylvain  and Piwowarski, Benjamin  and Staiano, Jacopo},
    booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
    year = {2020}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/mlsum_es

  • Config description: MLSum is a large-scale multiLingual summarization dataset. It is buillt from online news outlets, this split focusing on Spanish.

  • Download size: 501.27 MiB

  • Dataset size: 1.29 GiB

  • Auto-cached (documentation): No

  • Splits:

Split Examples
'challenge_test_covid' 1,938
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 13,366
'train' 259,888
'validation' 9,977
  • Feature structure:
FeaturesDict({
    'date': string,
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'target': string,
    'text': string,
    'title': string,
    'topic': string,
    'url': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
date Tensor string
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
target Tensor string
text Tensor string
title Tensor string
topic Tensor string
url Tensor string
  • Citation:
@inproceedings{scialom-etal-2020-mlsum,
    title = "{MLSUM}: The Multilingual Summarization Corpus",
    author = {Scialom, Thomas  and Dray, Paul-Alexis  and Lamprier, Sylvain  and Piwowarski, Benjamin  and Staiano, Jacopo},
    booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
    year = {2020}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/schema_guided_dialog

  • Config description: The Schema-Guided Dialogue (SGD) dataset contains 18K multi-domain task-oriented dialogues between a human and a virtual assistant, which covers 17 domains ranging from banks and events to media, calendar, travel, and weather.

  • Download size: 17.00 MiB

  • Dataset size: 201.19 MiB

  • Auto-cached (documentation): Yes (challenge_test_backtranslation, challenge_test_bfp02, challenge_test_bfp05, challenge_test_nopunc, challenge_test_scramble, challenge_train_sample, challenge_validation_sample, test, validation), Only when shuffle_files=False (train)

  • Splits:

Split Examples
'challenge_test_backtranslation' 500
'challenge_test_bfp02' 500
'challenge_test_bfp05' 500
'challenge_test_nopunc' 500
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 10,000
'train' 164,982
'validation' 10,000
  • Feature structure:
FeaturesDict({
    'context': Sequence(string),
    'dialog_acts': Sequence({
        'act': ClassLabel(shape=(), dtype=int64, num_classes=18),
        'slot': string,
        'values': Sequence(string),
    }),
    'dialog_id': string,
    'gem_id': string,
    'gem_parent_id': string,
    'prompt': string,
    'references': Sequence(string),
    'service': string,
    'target': string,
    'turn_id': int32,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
context Sequence(Tensor) (None,) string
dialog_acts Sequence
dialog_acts/act ClassLabel int64
dialog_acts/slot Tensor string
dialog_acts/values Sequence(Tensor) (None,) string
dialog_id Tensor string
gem_id Tensor string
gem_parent_id Tensor string
prompt Tensor string
references Sequence(Tensor) (None,) string
service Tensor string
target Tensor string
turn_id Tensor int32
  • Citation:
@article{rastogi2019towards,
  title={Towards Scalable Multi-domain Conversational Agents: The Schema-Guided Dialogue Dataset},
  author={Rastogi, Abhinav and Zang, Xiaoxue and Sunkara, Srinivas and Gupta, Raghav and Khaitan, Pranav},
  journal={arXiv preprint arXiv:1909.05855},
  year={2019}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/totto

  • Config description: ToTTo is a Table-to-Text NLG task. The task is as follows: Given a Wikipedia table with row names, column names and table cells, with a subset of cells highlighted, generate a natural language description for the highlighted part of the table.

  • Download size: 180.75 MiB

  • Dataset size: 645.86 MiB

  • Auto-cached (documentation): No

  • Splits:

Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 7,700
'train' 121,153
'validation' 7,700
  • Feature structure:
FeaturesDict({
    'example_id': string,
    'gem_id': string,
    'gem_parent_id': string,
    'highlighted_cells': Sequence(Sequence(int32)),
    'overlap_subset': string,
    'references': Sequence(string),
    'sentence_annotations': Sequence({
        'final_sentence': string,
        'original_sentence': string,
        'sentence_after_ambiguity': string,
        'sentence_after_deletion': string,
    }),
    'table': Sequence(Sequence({
        'column_span': int32,
        'is_header': bool,
        'row_span': int32,
        'value': string,
    })),
    'table_page_title': string,
    'table_section_text': string,
    'table_section_title': string,
    'table_webpage_url': string,
    'target': string,
    'totto_id': int32,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
example_id Tensor string
gem_id Tensor string
gem_parent_id Tensor string
highlighted_cells Sequence(Sequence(Tensor)) (None, None) int32
overlap_subset Tensor string
references Sequence(Tensor) (None,) string
sentence_annotations Sequence
sentence_annotations/final_sentence Tensor string
sentence_annotations/original_sentence Tensor string
sentence_annotations/sentence_after_ambiguity Tensor string
sentence_annotations/sentence_after_deletion Tensor string
table Sequence
table/column_span Tensor int32
table/is_header Tensor bool
table/row_span Tensor int32
table/value Tensor string
table_page_title Tensor string
table_section_text Tensor string
table_section_title Tensor string
table_webpage_url Tensor string
target Tensor string
totto_id Tensor int32
  • Citation:
@inproceedings{parikh2020totto,
  title=ToTTo: A Controlled Table-To-Text Generation Dataset,
  author={Parikh, Ankur and Wang, Xuezhi and Gehrmann, Sebastian and Faruqui, Manaal and Dhingra, Bhuwan and Yang, Diyi and Das, Dipanjan},
  booktitle={Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
  pages={1173--1186},
  year={2020}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/web_nlg_en

  • Config description: WebNLG is a bi-lingual dataset (English, Russian) of parallel DBpedia triple sets and short texts that cover about 450 different DBpedia properties. The WebNLG data was originally created to promote the development of RDF verbalisers able to generate short text and to handle micro-planning.

  • Download size: 12.57 MiB

  • Dataset size: 19.91 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'challenge_test_numbers' 500
'challenge_test_scramble' 500
'challenge_train_sample' 502
'challenge_validation_sample' 499
'test' 1,779
'train' 35,426
'validation' 1,667
  • Feature structure:
FeaturesDict({
    'category': string,
    'gem_id': string,
    'gem_parent_id': string,
    'input': Sequence(string),
    'references': Sequence(string),
    'target': string,
    'webnlg_id': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
category Tensor string
gem_id Tensor string
gem_parent_id Tensor string
input Sequence(Tensor) (None,) string
references Sequence(Tensor) (None,) string
target Tensor string
webnlg_id Tensor string
  • Citation:
@inproceedings{gardent2017creating,
  author = "Gardent, Claire
    and Shimorina, Anastasia
    and Narayan, Shashi
    and Perez-Beltrachini, Laura",
  title = "Creating Training Corpora for NLG Micro-Planners",
  booktitle = "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
  year = "2017",
  publisher = "Association for Computational Linguistics",
  pages = "179--188",
  location = "Vancouver, Canada",
  doi = "10.18653/v1/P17-1017",
  url = "http://www.aclweb.org/anthology/P17-1017"
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/web_nlg_ru

  • Config description: WebNLG is a bi-lingual dataset (English, Russian) of parallel DBpedia triple sets and short texts that cover about 450 different DBpedia properties. The WebNLG data was originally created to promote the development of RDF verbalisers able to generate short text and to handle micro-planning.

  • Download size: 7.49 MiB

  • Dataset size: 11.30 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'challenge_test_scramble' 500
'challenge_train_sample' 501
'challenge_validation_sample' 500
'test' 1,102
'train' 14,630
'validation' 790
  • Feature structure:
FeaturesDict({
    'category': string,
    'gem_id': string,
    'gem_parent_id': string,
    'input': Sequence(string),
    'references': Sequence(string),
    'target': string,
    'webnlg_id': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
category Tensor string
gem_id Tensor string
gem_parent_id Tensor string
input Sequence(Tensor) (None,) string
references Sequence(Tensor) (None,) string
target Tensor string
webnlg_id Tensor string
  • Citation:
@inproceedings{gardent2017creating,
  author = "Gardent, Claire
    and Shimorina, Anastasia
    and Narayan, Shashi
    and Perez-Beltrachini, Laura",
  title = "Creating Training Corpora for NLG Micro-Planners",
  booktitle = "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
  year = "2017",
  publisher = "Association for Computational Linguistics",
  pages = "179--188",
  location = "Vancouver, Canada",
  doi = "10.18653/v1/P17-1017",
  url = "http://www.aclweb.org/anthology/P17-1017"
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/wiki_auto_asset_turk

  • Config description: WikiAuto provides a set of aligned sentences from English Wikipedia and Simple English Wikipedia as a resource to train sentence simplification systems. ASSET and TURK are high-quality simplification datasets used for testing.

  • Download size: 121.01 MiB

  • Dataset size: 202.40 MiB

  • Auto-cached (documentation): Yes (challenge_test_asset_backtranslation, challenge_test_asset_bfp02, challenge_test_asset_bfp05, challenge_test_asset_nopunc, challenge_test_turk_backtranslation, challenge_test_turk_bfp02, challenge_test_turk_bfp05, challenge_test_turk_nopunc, challenge_train_sample, challenge_validation_sample, test_asset, test_turk, validation), Only when shuffle_files=False (train)

  • Splits:

Split Examples
'challenge_test_asset_backtranslation' 359
'challenge_test_asset_bfp02' 359
'challenge_test_asset_bfp05' 359
'challenge_test_asset_nopunc' 359
'challenge_test_turk_backtranslation' 359
'challenge_test_turk_bfp02' 359
'challenge_test_turk_bfp05' 359
'challenge_test_turk_nopunc' 359
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test_asset' 359
'test_turk' 359
'train' 483,801
'validation' 20,000
  • Feature structure:
FeaturesDict({
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'source': string,
    'target': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
source Tensor string
target Tensor string
  • Citation:
@inproceedings{jiang-etal-2020-neural,
    title = "Neural {CRF} Model for Sentence Alignment in Text Simplification",
    author = "Jiang, Chao  and
      Maddela, Mounica  and
      Lan, Wuwei  and
      Zhong, Yang  and
      Xu, Wei",
    booktitle = "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
    month = jul,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2020.acl-main.709",
    doi = "10.18653/v1/2020.acl-main.709",
    pages = "7943--7960",
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/xsum

  • Config description: The dataset is for the task of abstractive summarization in its extreme form, its about summarizing a document in a single sentence.

  • Download size: 246.31 MiB

  • Dataset size: 78.89 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'challenge_test_backtranslation' 500
'challenge_test_bfp_02' 500
'challenge_test_bfp_05' 500
'challenge_test_covid' 401
'challenge_test_nopunc' 500
'challenge_train_sample' 500
'challenge_validation_sample' 500
'test' 1,166
'train' 23,206
'validation' 1,117
  • Feature structure:
FeaturesDict({
    'document': string,
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'target': string,
    'xsum_id': string,
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
document Tensor string
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
target Tensor string
xsum_id Tensor string
  • Citation:
@inproceedings{Narayan2018dont,
  author = "Shashi Narayan and Shay B. Cohen and Mirella Lapata",
  title = "Don't Give Me the Details, Just the Summary! {T}opic-Aware Convolutional Neural Networks for Extreme Summarization",
  booktitle = "Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing ",
  year = "2018",
  address = "Brussels, Belgium",
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/wiki_lingua_arabic_ar

  • Config description: Wikilingua is a large-scale, multilingual dataset for the evaluation of cross-lingual abstractive summarization systems..

  • Download size: 56.25 MiB

  • Dataset size: 291.42 MiB

  • Auto-cached (documentation): No

  • Splits:

Split Examples
'test' 5,841
'train' 20,441
'validation' 2,919
  • Feature structure:
FeaturesDict({
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'source': string,
    'source_aligned': Translation({
        'ar': Text(shape=(), dtype=string),
        'en': Text(shape=(), dtype=string),
    }),
    'target': string,
    'target_aligned': Translation({
        'ar': Text(shape=(), dtype=string),
        'en': Text(shape=(), dtype=string),
    }),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
source Tensor string
source_aligned Translation
source_aligned/ar Text string
source_aligned/en Text string
target Tensor string
target_aligned Translation
target_aligned/ar Text string
target_aligned/en Text string
  • Citation:
@inproceedings{ladhak-wiki-2020,
title=WikiLingua: A New Benchmark Dataset for Multilingual Abstractive Summarization,
author={Faisal Ladhak, Esin Durmus, Claire Cardie and Kathleen McKeown},
booktitle={Findings of EMNLP, 2020},
year={2020}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/wiki_lingua_chinese_zh

  • Config description: Wikilingua is a large-scale, multilingual dataset for the evaluation of cross-lingual abstractive summarization systems..

  • Download size: 31.38 MiB

  • Dataset size: 122.06 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'test' 3,775
'train' 13,211
'validation' 1,886
  • Feature structure:
FeaturesDict({
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'source': string,
    'source_aligned': Translation({
        'en': Text(shape=(), dtype=string),
        'zh': Text(shape=(), dtype=string),
    }),
    'target': string,
    'target_aligned': Translation({
        'en': Text(shape=(), dtype=string),
        'zh': Text(shape=(), dtype=string),
    }),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
source Tensor string
source_aligned Translation
source_aligned/en Text string
source_aligned/zh Text string
target Tensor string
target_aligned Translation
target_aligned/en Text string
target_aligned/zh Text string
  • Citation:
@inproceedings{ladhak-wiki-2020,
title=WikiLingua: A New Benchmark Dataset for Multilingual Abstractive Summarization,
author={Faisal Ladhak, Esin Durmus, Claire Cardie and Kathleen McKeown},
booktitle={Findings of EMNLP, 2020},
year={2020}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/wiki_lingua_czech_cs

  • Config description: Wikilingua is a large-scale, multilingual dataset for the evaluation of cross-lingual abstractive summarization systems..

  • Download size: 13.84 MiB

  • Dataset size: 58.05 MiB

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'test' 1,438
'train' 5,033
'validation' 718
  • Feature structure:
FeaturesDict({
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'source': string,
    'source_aligned': Translation({
        'cs': Text(shape=(), dtype=string),
        'en': Text(shape=(), dtype=string),
    }),
    'target': string,
    'target_aligned': Translation({
        'cs': Text(shape=(), dtype=string),
        'en': Text(shape=(), dtype=string),
    }),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
source Tensor string
source_aligned Translation
source_aligned/cs Text string
source_aligned/en Text string
target Tensor string
target_aligned Translation
target_aligned/cs Text string
target_aligned/en Text string
  • Citation:
@inproceedings{ladhak-wiki-2020,
title=WikiLingua: A New Benchmark Dataset for Multilingual Abstractive Summarization,
author={Faisal Ladhak, Esin Durmus, Claire Cardie and Kathleen McKeown},
booktitle={Findings of EMNLP, 2020},
year={2020}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/wiki_lingua_dutch_nl

  • Config description: Wikilingua is a large-scale, multilingual dataset for the evaluation of cross-lingual abstractive summarization systems..

  • Download size: 53.88 MiB

  • Dataset size: 237.97 MiB

  • Auto-cached (documentation): Yes (test, validation), Only when shuffle_files=False (train)

  • Splits:

Split Examples
'test' 6,248
'train' 21,866
'validation' 3,123
  • Feature structure:
FeaturesDict({
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'source': string,
    'source_aligned': Translation({
        'en': Text(shape=(), dtype=string),
        'nl': Text(shape=(), dtype=string),
    }),
    'target': string,
    'target_aligned': Translation({
        'en': Text(shape=(), dtype=string),
        'nl': Text(shape=(), dtype=string),
    }),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
source Tensor string
source_aligned Translation
source_aligned/en Text string
source_aligned/nl Text string
target Tensor string
target_aligned Translation
target_aligned/en Text string
target_aligned/nl Text string
  • Citation:
@inproceedings{ladhak-wiki-2020,
title=WikiLingua: A New Benchmark Dataset for Multilingual Abstractive Summarization,
author={Faisal Ladhak, Esin Durmus, Claire Cardie and Kathleen McKeown},
booktitle={Findings of EMNLP, 2020},
year={2020}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/wiki_lingua_english_en

  • Config description: Wikilingua is a large-scale, multilingual dataset for the evaluation of cross-lingual abstractive summarization systems..

  • Download size: 112.56 MiB

  • Dataset size: 657.51 MiB

  • Auto-cached (documentation): No

  • Splits:

Split Examples
'test' 28,614
'train' 99,020
'validation' 13,823
  • Feature structure:
FeaturesDict({
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'source': string,
    'source_aligned': Translation({
        'en': Text(shape=(), dtype=string),
    }),
    'target': string,
    'target_aligned': Translation({
        'en': Text(shape=(), dtype=string),
    }),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
source Tensor string
source_aligned Translation
source_aligned/en Text string
target Tensor string
target_aligned Translation
target_aligned/en Text string
  • Citation:
@inproceedings{ladhak-wiki-2020,
title=WikiLingua: A New Benchmark Dataset for Multilingual Abstractive Summarization,
author={Faisal Ladhak, Esin Durmus, Claire Cardie and Kathleen McKeown},
booktitle={Findings of EMNLP, 2020},
year={2020}
}
@article{gehrmann2021gem,
  author    = {Sebastian Gehrmann and
               Tosin P. Adewumi and
               Karmanya Aggarwal and
               Pawan Sasanka Ammanamanchi and
               Aremu Anuoluwapo and
               Antoine Bosselut and
               Khyathi Raghavi Chandu and
               Miruna{-}Adriana Clinciu and
               Dipanjan Das and
               Kaustubh D. Dhole and
               Wanyu Du and
               Esin Durmus and
               Ondrej Dusek and
               Chris Emezue and
               Varun Gangal and
               Cristina Garbacea and
               Tatsunori Hashimoto and
               Yufang Hou and
               Yacine Jernite and
               Harsh Jhamtani and
               Yangfeng Ji and
               Shailza Jolly and
               Dhruv Kumar and
               Faisal Ladhak and
               Aman Madaan and
               Mounica Maddela and
               Khyati Mahajan and
               Saad Mahamood and
               Bodhisattwa Prasad Majumder and
               Pedro Henrique Martins and
               Angelina McMillan{-}Major and
               Simon Mille and
               Emiel van Miltenburg and
               Moin Nadeem and
               Shashi Narayan and
               Vitaly Nikolaev and
               Rubungo Andre Niyongabo and
               Salomey Osei and
               Ankur P. Parikh and
               Laura Perez{-}Beltrachini and
               Niranjan Ramesh Rao and
               Vikas Raunak and
               Juan Diego Rodriguez and
               Sashank Santhanam and
               Jo{\~{a} }o Sedoc and
               Thibault Sellam and
               Samira Shaikh and
               Anastasia Shimorina and
               Marco Antonio Sobrevilla Cabezudo and
               Hendrik Strobelt and
               Nishant Subramani and
               Wei Xu and
               Diyi Yang and
               Akhila Yerukola and
               Jiawei Zhou},
  title     = {The {GEM} Benchmark: Natural Language Generation, its Evaluation and
               Metrics},
  journal   = {CoRR},
  volume    = {abs/2102.01672},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.01672},
  archivePrefix = {arXiv},
  eprint    = {2102.01672}
}

Note that each GEM dataset has its own citation. Please see the source to see
the correct citation for each contained dataset."

gem/wiki_lingua_french_fr

  • Config description: Wikilingua is a large-scale, multilingual dataset for the evaluation of cross-lingual abstractive summarization systems..

  • Download size: 113.26 MiB

  • Dataset size: 522.28 MiB

  • Auto-cached (documentation): No

  • Splits:

Split Examples
'test' 12,731
'train' 44,556
'validation' 6,364
  • Feature structure:
FeaturesDict({
    'gem_id': string,
    'gem_parent_id': string,
    'references': Sequence(string),
    'source': string,
    'source_aligned': Translation({
        'en': Text(shape=(), dtype=string),
        'fr': Text(shape=(), dtype=string),
    }),
    'target': string,
    'target_aligned': Translation({
        'en': Text(shape=(), dtype=string),
        'fr': Text(shape=(), dtype=string),
    }),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
gem_id Tensor string
gem_parent_id Tensor string
references Sequence(Tensor) (None,) string
source Tensor string
source_aligned Translation
source_aligned/en Text string
source_aligned/fr Text string
target Tensor string
target_aligned Translation
target_aligned/en Text string
target_aligned/fr Text string