Skip to content

Dataset · ucsbai

LIAR fake news detection and fact-checking dataset

ucsbai/liar

12.8 thousand short political statements with truthfulness labels and editorial explanations.

What's inside

The corpus contains 12.8 thousand short English statements collected via the politifact.com API. Each claim is manually verified by an editor, assigned a truthfulness category, and accompanied by a detailed analytical justification of the assessment.

What you can do with it

  • Training classifiers for automated fact-checking and fake news detection
  • Developing models that generate textual justifications for assessing credibility
Tasks
Text classification
Languages
English
Size
10K–100K rows
Downloads this month
1.6K
Likes
33

How to load it

from datasets import load_dataset

ds = load_dataset("ucsbai/liar")