VoiceStick: A French Spontaneous Speech Corpus for Voice-Guided Drone Teleoperation

VoiceStick is the first French spontaneous speech corpus dedicated to voice-guided drone teleoperation. It was collected from 29 dyads of native French-speaking participants using an asymmetric guide-pilot paradigm in a mixed-reality setting. The corpus captures the natural dynamics of spontaneous vocal interaction in a real-time teleoperation task, including hesitations, reformulations, and prosodic variability. The corpus comprises 4,219 utterances totaling 19,829 words, with a vocabulary of 669 unique words and 2,421 distinct spoken commands. If you use this corpus, please cite: Henry, A., Rossato, S., Graff, C., Gomez-Balderas, J.-E., & Huet, S. (2026). VoiceStick: A Spontaneous Speech Corpus for Drone Voice Guidance. *Proceedings of CORIA-TALN 2026*. Nantes, France.