Parallel English–Akan Navigation Dataset

It contains a curated bilingual parallel corpus developed to support domain-specific speech and language technologies, primarily, automatic speech recognition (ASR) and text-to-speech (TTS). The corpus was constructed to address the scarcity of navigation and accessibility-specific data for low-resource African languages, particularly Akan, and to support voice-enabled mobility aids for visually impaired users.

It comprises 6,600 English–Akan parallel sentences spanning seven functional navigation categories: route instructions, hazard and safety alerts, contextual orientation cues, status and confirmation feedback, landmark and location references, environmental descriptions, and system status prompts. Navigation topics represented include turn-by-turn route guidance, gutter and vehicle hazards, road crossing, cardinal direction and spatial positioning, terrain and elevation cues, on-track and off-track confirmations, campus halls and academic buildings, gates and junctions, indoor and outdoor environmental description, and device status and connectivity prompts.