annotations_creators:
- found
language_creators:
- found
language:
- en
license:
- mit
multilinguality:
- monolingual
size_categories:
- 100K<n<1M
source_datasets:
- extended|gigaword_2003
task_categories:
- summarization
task_ids: []
pretty_name: Gigaword
tags:
- headline-generation
dataset_info:
features:
- name: document
dtype: string
- name: summary
dtype: string
splits:
- name: train
num_bytes: 915246340
num_examples: 3803957
- name: validation
num_bytes: 45766944
num_examples: 189651
- name: test
num_bytes: 450774
num_examples: 1951
download_size: 578402958
dataset_size: 961464058
train-eval-index:
- config: default
task: summarization
task_id: summarization
splits:
train_split: train
eval_split: test
col_mapping:
document: text
summary: target
metrics:
- type: rouge
name: Rouge
Dataset Card for Gigaword
Table of Contents
- Dataset Description
- Dataset Structure
- Dataset Creation
- Considerations for Using the Data
- Additional Information
Dataset Description
- Repository: Gigaword repository
- Leaderboard: Gigaword leaderboard
- Paper: A Neural Attention Model for Abstractive Sentence Summarization
- Point of Contact: Alexander Rush
- Size of downloaded dataset files: 578.41 MB
- Size of the generated dataset: 962.96 MB
- Total amount of disk used: 1.54 GB
Dataset Summary
Headline-generation on a corpus of article pairs from Gigaword consisting of around 4 million articles. Use the 'org_data' provided by https://github.com/microsoft/unilm/ which is identical to https://github.com/harvardnlp/sent-summary but with better format.
Supported Tasks and Leaderboards
summarization: This dataset can be used for Summarization, where given a dicument, the goal is to predict its summery. The model performance is evaluated using the ROUGE metric. The leaderboard for this task is available here.
Languages
English.
Dataset Structure
Data Instances
An example of 'train' looks as follows.
{
'document': "australia 's current account deficit shrunk by a record #.## billion dollars -lrb- #.## billion us -rrb- in the june quarter due to soaring commodity prices , figures released monday showed .",
'summary': 'australian current account deficit narrows sharply'
}
Data Fields
The data fields are the same among all splits.
document: astringfeature.summary: astringfeature.
Data Splits
| name | train | validation | test |
|---|---|---|---|
| default | 3803957 | 189651 | 1951 |
Dataset Creation
Curation Rationale
Source Data
Initial Data Collection and Normalization
From the paper:
For our training set, we pair the headline of each article with its first sentence to create an inputsummary pair. While the model could in theory be trained on any pair, Gigaword contains many spurious headline-article pairs. We therefore prune training based on the following heuristic filters: (1) Are there no non-stop-words in common? (2) Does the title contain a byline or other extraneous editing marks? (3) Does the title have a question mark or colon? After applying these filters, the training set consists of roughly J = 4 million title-article pairs. We apply a minimal preprocessing step using PTB tokenization, lower-casing, replacing all digit characters with #, and replacing of word types seen less than 5 times with UNK. We also remove all articles from the time-period of the DUC evaluation. release. The complete input training vocabulary consists of 119 million word tokens and 110K unique word types with an average sentence size of 31.3 words. The headline vocabulary consists of 31 million tokens and 69K word types with the average title of length 8.3 words (note that this is significantly shorter than the DUC summaries). On average there are 4.6 overlapping word types between the headline and the input; although only 2.6 in the first 75-characters of the input.
Who are the source language producers?
From the paper:
For training data for both tasks, we utilize the annotated Gigaword data set (Graff et al., 2003; Napoles et al., 2012), which consists of standard Gigaword, preprocessed with Stanford CoreNLP tools (Manning et al., 2014).
Annotations
Annotation process
Annotations are inherited from the annotatated Gigaword data set.
Additional information from the paper:
Our model only uses annotations for tokenization and sentence separation, although several of the baselines use parsing and tagging as well.
Who are the annotators?
Personal and Sensitive Information
Considerations for Using the Data
Social Impact of Dataset
Discussion of Biases
Other Known Limitations
Additional Information
Dataset Curators
Licensing Information
Citation Information
@article{graff2003english,
title={English gigaword},
author={Graff, David and Kong, Junbo and Chen, Ke and Maeda, Kazuaki},
journal={Linguistic Data Consortium, Philadelphia},
volume={4},
number={1},
pages={34},
year={2003}
}
@article{Rush_2015,
title={A Neural Attention Model for Abstractive Sentence Summarization},
url={http://dx.doi.org/10.18653/v1/D15-1044},
DOI={10.18653/v1/d15-1044},
journal={Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing},
publisher={Association for Computational Linguistics},
author={Rush, Alexander M. and Chopra, Sumit and Weston, Jason},
year={2015}
}
Contributions
Thanks to @lewtun, @lhoestq, @thomwolf for adding this dataset.