Kasagadi: West African Radio Broadcast Fact-Check Dataset

Ghanaian and Nigerian radio broadcasts transcribed and fact-checked. The dataset contains approximately 74,500 segments from Twi-language broadcasts on Angel FM and Peace FM in Ghana and Hausa-language broadcasts from a Nigerian FM station, collected between March and June 2026. It was created by the African-Languages-Lab.

Use Cases

• Train fact-checking models based on AI-verified claims from radio broadcasts.
• Develop speech-to-text and NLP tools for Twi and Hausa languages based on transcribed radio content.
• Analyze misinformation trends in West African media based on fact-checked broadcast segments.
• Study linguistic patterns and code-switching in live radio broadcasts from Ghana and Nigeria.

Strengths

• Multilingual coverage of Twi and Hausa, which are major West African languages.
• Approximately 74,500 segments from over 480 total broadcasts, providing substantial textual data.
• Segments are AI fact-checked, adding a layer of verification to the content.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic and temporal bias inherent to the selected FM stations and the March–June 2026 period.

Provenance

Source

African-Languages-Lab

Collection Method

Transcribed, translated, and AI fact-checked segments from live FM radio broadcasts.

Time Range

Mar – Jun 2026

Freshness

Last updated 2026-06-11 13:17:45; freshness should be verified.

Geography

Ghana and Nigeria