Fact-Checking Dataset for African Languages

AfrIFact is a dataset for cultural information retrieval, evidence extraction, and fact-checking in African languages. It was created by Masakhane and last updated on April 2, 2026. The dataset is designed to assess the veracity of online claims, particularly those concerning healthcare and culture in low-resource linguistic contexts.

Use Cases

• Train models to classify claim_veracity based on extracted evidence text.
• Perform cross-lingual information retrieval using claim_text and retrieved_document fields.
• Develop evidence extraction models to identify supporting or refuting passages from source documents.
• Benchmark multilingual natural language inference systems on African language pairs.
• Analyze cultural or healthcare-related misinformation patterns across different languages.

Strengths

• Covers multiple African languages, addressing a low-resource area.
• Includes tasks for information retrieval, evidence extraction, and fact-checking.

Limitations

• Specific row count, column details, and sample size are unknown.
• Potential geographic and linguistic bias based on the subset of African languages included.
• Temporal coverage and data freshness for the claims are unspecified.

Provenance

Source

Masakhane

Collection Method

null

Time Range

null

Freshness

Last updated on 2026-04-02.

Geography

Africa