Datasets:
metadata
task_categories:
- summarization
language:
- en
CNN/DailyMail dataset filtered by using the code by Aumiller et al. (1) available at https://github.com/dennlinger/summaries/tree/main
min_length_summary = 18; min_length_reference = 150; length_metric = "whitespace"
bi-gram_overlap_fraction between summary and original text <= 0.630
min_compression_ratio = 2.5
Moreover: - all articles longer than 512 tokens were removed from the dataset. - Each line break ("\n") inside of a highlight was replaced by an empty space (" ").
(1): Aumiller, D., Fan, J., & Gertz, M. (2023). On the State of German (Abstractive) Text Summarization. arXiv preprint arXiv:2301.07095.