walzen/

gigaword

Putting it here for better accessibility.

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. There are two features:

document: article. summary: headline.

Homepage: https://github.com/harvardnlp/sent-summary Original Source code:… See the full description on the dataset page: https://huggingface.co/datasets/walzen/gigaword.