Technology

Our Technology in Action

Our speech recognition API handles accents, background noises, and the natural unpredictability of kids’ speech. Percentage based scores are returned for target phrases, individual words, and for every speech sound in the target phrase, enabling scoring down to the phoneme level.
 
To use the demo:
Choose a sentence from the drop down – press play – hear the audio – see our system return scores. Choose another sentence to try.

Our Technology is State-of-the-Art

SoapBox Labs builds speech recognition technology for kids using state-of-the-art, deep neural network machine learning and our proprietary, real-world kids’ speech dataset. Partners can integrate either our online or offline solution, both of which use our proprietary models to offer the same high level of accuracy and performance.

 

Online 

SaaS/Cloud

Available as a Cloud based RESTful Web Service, deployed globally in multiple data centers and configurable to incorporate custom/bespoke language models built according to specific customer requirements.

Offline

On Device

Available on both Android and IOS devices. Built for ArmV8-A, our technology is available as static and dynamic libraries, and potentially also via a binary. SDKs are available in a variety of languages. The offline product is also configurable by age and can be tailored to include bespoke acoustic & language models upon request.

Offline

Embedded

(in development)

Deployment of our deep learning models for embedded offline ultra low power chips and neural processors/inference engines.

Test the world’s most accurate kids speech recognition technology with your kids’ products today.