metadata

annotations_creators:
  - machine-generated
language_creators:
  - machine-generated
languages:
  - en
licenses:
  - cc-by-nc-sa-4.0
multilinguality:
  - monolingual
pretty_name: rebel-dataset
size_categories:
  - unknown
source_datasets:
  - original
task_categories:
  - text-retrieval
  - conditional-text-generation
task_ids:
  - text-retrieval-other-relation-extraction

Dataset Card for REBEL dataset

Dataset Card for REBEL dataset

Dataset Description

Repository: https://github.com/Babelscape/rebel
Paper: https://github.com/Babelscape/rebel/blob/main/docs/EMNLP_2021_REBEL__Camera_Ready_.pdf
Point of Contact: huguetcabot@babelscape.com

Dataset Summary

Dataset created for REBEL dataset from interlinking Wikidata and Wikipedia for Relation Extraction, filtered using NLI.

Supported Tasks and Leaderboards

text-retrieval-other-relation-extraction: The dataset can be used to train a model for Relation Extraction, which consists in extracting triplets from raw text, made of subject, object and relation type. Success on this task is typically measured by achieving a high/low F1. The BART) model currently achieves the following score. [IF A LEADERBOARD IS AVAILABLE]: 74.

Languages

The dataset is in English, from the English Wikipedia.

Dataset Structure

Data Instances

Provide an JSON-formatted example and brief description of a typical instance in the dataset. If available, provide a link to further examples.

{
  'example_field': ...,
  ...
}

Provide any additional information that is not covered in the other sections about the data here. In particular describe any relationships between data points and if these relationships are made explicit.

Data Fields

List and describe the fields present in the dataset. Mention their data type, and whether they are used as input or output in any of the tasks the dataset currently supports. If the data has span indices, describe their attributes, such as whether they are at the character level or word level, whether they are contiguous or not, etc. If the datasets contains example IDs, state whether they have an inherent meaning, such as a mapping to other datasets or pointing to relationships between data points.

example_field: description of example_field

Note that the descriptions can be initialized with the Show Markdown Data Fields output of the tagging app, you will then only need to refine the generated descriptions.

Data Splits

Describe and name the splits in the dataset if there are more than one.

Describe any criteria for splitting the data, if used. If their 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.

Provide the sizes of each split. As appropriate, provide any descriptive statistics for the features, such as average length. For example:

	Tain	Valid	Test
Input Sentences
Average Sentence Length

Dataset Creation

Curation Rationale

This dataset was created to enable the training of a BART based model as pre-training phase for Relation Extraction as seen in the paper REBEL: Relation Extraction By End-to-end Language generation.

Source Data

Data comes from Wikipedia text before the table of contents, as well as Wikidata for the triplets annotation.

Initial Data Collection and Normalization

For the data collection, the dataset extraction pipeline cRocoDiLe: Automatic Relation Extraction Dataset with NLI filtering insipired by T-REx Pipeline more details found at: T-REx Website. The starting point is a Wikipedia dump as well as a Wikidata one.

After the triplets are extracted, an NLI system was used to filter out those not entailed by the text.

Who are the source language producers?

Any Wikipedia and Wikidata contributor.

Annotations

Annotation process

TThe dataset extraction pipeline cRocoDiLe: Automatic Relation Extraction Dataset with NLI filtering.

Who are the annotators?

Automatic annottations

Personal and Sensitive Information

All text is from Wikipedia, any Personal or Sensitive Information there may be present in this dataset.

Considerations for Using the Data

Social Impact of Dataset

The dataset serves as a pre-training step for Relation Extraction models. It is distantly annotated, hence it should only be used as such. A model trained solely on this dataset may produce allucinations coming from the silver nature of the dataset.

Discussion of Biases

Since the dataset was automatically created from Wikipedia and Wikidata, it may reflect the biases withing those sources.

For Wikipedia text, see for example Dinan et al 2020 on biases in Wikipedia (esp. Table 1), or Blodgett et al 2020 for a more general discussion of the topic.

For Wikidata, there are class imbalances, also resulting from Wikipedia.

Other Known Limitations

Not for now

Additional Information

Dataset Curators

Pere-Lluis Huguet Cabot - Babelscape and Sapienza University of Rome, Italy Roberto Navigli - Sapienza University of Rome, Italy

Licensing Information

Contents of this repository are restricted to only non-commercial research purposes under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). Copyright of the dataset contents belongs to the original copyright holders.

Citation Information

Provide the BibTex-formatted reference for the dataset. For example:

@inproceedings{huguet-cabot-navigli-2021-rebel,
title = "REBEL: Relation Extraction By End-to-end Language generation",
author = "Huguet Cabot, Pere-Llu{\'\i}s  and
    Navigli, Roberto",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2021",
month = nov,
year = "2021",
address = "Online and in the Barceló Bávaro Convention Centre, Punta Cana, Dominican Republic",
publisher = "Association for Computational Linguistics",
url = "https://github.com/Babelscape/rebel/blob/main/docs/EMNLP_2021_REBEL__Camera_Ready_.pdf",
}

Contributions

Thanks to @littlepea13 for adding this dataset.

Datasets:

Babelscape
/

rebel-dataset