Test split of the Newsroom dataset (2) filtered by using the code by Aumiller et al. (1) available at
https://github.com/dennlinger/summaries/tree/main with the following settings:
min_length_summary = 18; min_length_reference = 250; length_metric = "whitespace"
Furthermore:
all article-summary pairs with density > 2 were removed. This means that all summaries contained in the dataset are more on the abstractive side
each line break ("\n") contained in an article or a summary was replaced… See the full description on the dataset page:
https://huggingface.co/datasets/giuliadc/newsroom_test_abstr_no-line-breaks.
