AfriHumor is an annotator-driven workbench for building a humor detection and generation dataset — African languages plus English and French. We capture word-level speech, prosody, visual cues, and linguistic structure from comedy — with per-language tonal and diacritic input support where needed. We start with the languages below and expand coverage as new annotators come on board.
Coverage grows with the team — new languages open up as annotators join.
Each comedy clip is annotated across four modalities so the dataset supports both humor detection and generation — without relying on laughter markers.