Case Study: Largest Software and Technology Company achieves more accurate multilingual voice search with Lionbridge

How We Help Train Your Machines To Understand Multilingual Voice Search

Largest Software and Technology Company, a global leader that builds a multilingual search engine and voice-activated personal assistant, faced the challenge of understanding voice queries from diverse speakers, accents, ages, devices and noisy environments worldwide. To train and validate its speech-recognition models, the company engaged Lionbridge for speech data collection, creation and testing services.

Lionbridge provided an end-to-end solution—studio-quality recordings, crowd-sourced speech samples via its 400,000+ global network, and 240 hours of ambient-noise capture—collecting over 20 hours of speech across 30+ languages and speakers aged 6–75. With Lionbridge’s dataset, the customer can better train, calibrate and test recognition algorithms to produce more accurate voice-search results and correct actions across languages and environments, improving user experience and strengthening global brand loyalty.