Danish Fact-Checking Dataset for Text Classification

For fact-checking and natural language inference tasks in the Danish language. It contains between 1,000 and 10,000 rows, as indicated by its size category, and features expert-generated annotations. The data was created by strombergnlp and last updated in October 2022.

Use Cases

• Train a text classification model for fact-checking using the expert-generated annotations.
• Perform natural language inference (NLI) analysis on Danish language claims.
• Analyze the distribution of verified claims within the dataset's 1K-10K sample size.

Strengths

• Contains between 1,000 and 10,000 data points, as per its size category.
• Annotations are expert-generated, suggesting a degree of label reliability.
• Specifically targets the Danish language, providing a resource for a less common language in NLP.

Limitations

• The exact row count, column structure, and sample data are unknown.
• The dataset is relatively small, with a maximum of 10,000 rows, which may limit model training.
• The data is monolingual (Danish) and regionally focused (US), limiting generalizability.

Provenance

Source

Original source datasets, as indicated by tags.

Collection Method

Expert-generated annotations.

Freshness

Last updated on 2022-10-25.

Geography

Region: US (as per tags).