Skip to content

Dataset · webis

Reddit posts corpus with short TL;DR summaries

webis/tldr-17

Nearly 3.85 million English Reddit posts with author summaries for automatic summarization tasks.

What's inside

The dataset contains 3,848,330 preprocessed English Reddit posts. The average main text length is 270 words, and the short summary is 28 words. Besides text and summary, each record contains metadata about the author and subreddit.

What you can do with it

  • Training and fine-tuning abstractive text summarization models
  • Training systems to compress user content and news into a single paragraph
Tasks
Summarization
Languages
English
Size
1M–10M rows
Downloads this month
1.7K
Likes
59

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("webis/tldr-17")