We provide ASR models a variety of data services like reading, face-to-face conversation, multi-user dialogue and so on. These datasets and services cover all target languages, dialect variation, different background noise, genders, ages in real world scenarios. We have unique and strict criterion for speakers, recording environment, devices, annotation specification, acceptance inspection and so on in order to insure the uniformity, stability and reliable quality of the products and services.

We have delivered a lot of datasets that cover at least 50 languages (with different accents) and 20 dialects for diverse scenarios. These high-quality datasets make our customers’ ASR models be trained well and help them get notable improvement of the recognition accuracy. We assist our customers accelerating their go-to-market strategy and boost the smart home controls industry through our products and services.