Building a Diverse In-Car Hindi Speech Dataset.

Client's Challenge & Our Solution

A leading automotive technology client partnered with FutureBeeAI to collect diverse in-car speech data in Hindi, focusing on various states in India. The goal was to gather scripted phrases, wake words, and commands while capturing environmental variations like open and closed windows, indoor and outdoor parking, and the car's AC and engine on and off.

We utilized our speech data collection platform, Yugo, to onboard 450 participants from different Hindi-speaking states, ensuring representation across genders and age groups. This approach enabled us to create a comprehensive dataset that met the client’s specific diversity and environmental requirements.

Outcome & Features:

Successfully gathered in-car speech data from 450 participants across multiple states in India.

Captured environmental variations to enhance the dataset's real-world applicability.

Delivered data in a fully compliant, quality-controlled format that boosted the performance of the client's voice assistant by 30%.