Claim Normalization to Empower Fact-Checking

DATA DESCRIPTION

The dataset consists of fact-checked social media posts paired with concise, claim-oriented summaries (referred to as normalized claims), which are created by professional fact-checkers as part of the verification process. Because the dataset is not subjected to external annotation, it avoids potential biases and maintains a high level of quality.

It includes predefined training, development, and test splits, comprising nearly 6,000 posts in total.

The dataset is intentionally left unprocessed to better reflect real-world extraction scenarios. Additionally, no metadata is provided in order to preserve user privacy and protect personally identifiable information (PII).

ETHICAL USAGE COMMITMENT

By requesting access, users acknowledge and agree to use the dataset solely for ethical and lawful research purposes. The following uses are strictly prohibited:

Developing tools or algorithms intended to promote or assist in illicit activities or to evade law enforcement or regulatory oversight.

Any form of algorithmic discrimination or bias based on protected characteristics such as race, gender, sexual orientation, religion, or political beliefs.

Unauthorized access, data leakage, or any action that compromises the confidentiality or integrity of the dataset.