cnn_dailymail
Over 300k news articles for summarization and comprehension tasks.
Open dataset Updated 2026-06-18
What is cnn_dailymail?
The dataset consists of just over 300k unique news articles written by journalists at CNN and the Daily Mail. Current versions enable both extractive and abstractive summarization.
It is useful for training and evaluating models on summarization, machine reading comprehension, and question answering.
What you can build with cnn_dailymail
Training Abstractive Summarizers
Develop and fine-tune models like BART or T5 to generate concise summaries from full news articles using the provided highlights as targets.
Extractive Summarization Systems
Build models that select key sentences from CNN and Daily Mail articles to create summaries without generating new text.
News-Based QA Prototypes
Train reading comprehension models on the original CNN/DailyMail articles to answer questions derived from the article content.
Load cnn_dailymail
Python
from datasets import load_dataset
ds = load_dataset("abisee/cnn_dailymail")- 1pip install datasets
- 2from datasets import load_dataset
- 3ds = load_dataset('cnn_dailymail', '3.0.0')
- 4Access splits via ds['train'], ds['validation'], ds['test']
- 5Use 'article' and 'highlights' fields for training
cnn_dailymail: pros & cons
Pros
- +Large scale with over 300k articles
- +Supports both extractive and abstractive tasks
- +Widely used benchmark in summarization research
- +Easy loading via Hugging Face
Cons
- –Articles are from 2007-2015 period only
- –Some articles contain boilerplate text
- –Highlights are bullet-point style, not full paragraphs
Did you find this helpful?
Frequently asked questions
A collection of over 300,000 English news articles from CNN and Daily Mail originally created for summarization and reading comprehension tasks.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar datasets
Other text & nlp options worth comparing.