A Gender-Balanced and Region-Diverse Luganda Speech Dataset for Automatic Speech Recognition

This dataset contains a gender, age, and region-balanced Luganda speech corpus designed for bias-aware and fair Automatic Speech Recognition (ASR) research. The dataset includes speech recordings collected from diverse speakers across the Central, Eastern, Northern, and Western regions of Uganda. Each audio file is accompanied by demographic metadata including speaker gender, age group, and region of origin. The data has been curated and filtered to ensure equal representation of male and female speakers within every region–age group. The dataset is intended to support research on fairness, bias mitigation, and performance evaluation in Luganda ASR systems, as well as linguistic and sociophonetic analysis of regional and demographic variation in Luganda speech.