Data Collection
Multilingual Voice Recordings
TransPerfect supports client with large-scale, multilingual voice data collection to streamline their app localization process.
The Problem
There is an increasing demand for solutions that voice-controlled smart products can deliver. To meet the end user's expectations, voice identification must be accurate, regardless of background noises or language and voice parameters.
To prevent possible bias, you need a large-scale audio collection of native speakers of the target languages covering various demographics and environments. Our client, an industry leader in the far-field speech and voice recognition market, did not have the necessary internal resources to execute such a complex data collection. Therefore, they requested TransPerfect’s support in the Korean and Chinese markets.
• • • •The Solution• • • •
To expand the coverage of the machine-learning solution in Mandarin and Korean, TransPerfect resourced more than 500 participants from various demographic groups for each language. The participants were asked to complete ten recording sessions using a TransPerfect app on their mobile phone. These sessions were conducted at different locations and times of day to capture various background noises and voice parameters.
TransPerfect delivered the project in less than eight weeks. The audio data sets from the participants’ recordings enabled the improvement of the client’s audio/voice-recognition solutions.
DataForce has a global community of over 1,000,000 members from around the globe and linguistic experts in over 250 languages. DataForce is its own platform but can also use client or third-party tools. This way, your data is always under control.