Case Study: Far-field Speech and Voice Recognition Company achieves large-scale Mandarin and Korean audio collection and improved recognition accuracy with TransPerfect

TransPerfect supports client with large-scale, multilingual voice data collection to streamline their app localization process

Far-field Speech and Voice Recognition Company needed large-scale, unbiased multilingual audio data to improve far-field voice identification across background noises, languages and demographics but lacked the internal resources to collect it in key Asian markets. They engaged TransPerfect, using its DataForce community and mobile recording tools, to source targeted voice samples in Korean and Mandarin.

TransPerfect recruited more than 500 participants per language and ran ten mobile recording sessions per participant at varied times and locations to capture diverse noise and voice parameters, completing the project in under eight weeks. The resulting audio datasets fed into the customer’s machine‑learning pipeline and enabled measurable improvements to their audio/voice-recognition solutions, leveraging TransPerfect’s DataForce network of contributors and linguistic expertise.