ANSWERED: Kids Voice Tech in EdTech
June 3, 2021
Last week’s webinar, “Unlocking the Power of Voice Technology in Education,” was all about how voice technology is being used to support kids in learning to read, do math, and learn languages.
The Q&A at the end of the session ran over with questions, so here’s a review of five of them — and the answers too!
1. Does your voice tech check for gender or other forms of bias?
Yes. We do rigorous testing of the accuracy of our voice engine using voice data from a huge diversity of kids: different ages, locations, backgrounds, accents, dialects, and genders to ensure it works equally well across a range of metadata.
Importantly, our voice engine has also been independently verified by a number of our clients to show no bias in performance across gender, accents and dialects, and socioeconomic backgrounds. We’re more than a little proud of that.
2. Is it possible to assess a child’s pronunciation from spontaneous speech as well as from read speech?
Yes. Our voice engine offers a range of solutions including Verification and Fluency (assessment) of a target text. We also offer speech-to-text (STT), returning the transcription of the audio file as well as the confidence score for each word.
Also, we will soon be releasing a feature where the confidence scores of the individual phonemes are also returned.
3. Is SoapBox pursuing automatic scoring of prosody?
Yes! Many clients have asked for prosody, and we’ll be delivering it for them very shortly.
We’re currently looking at the best way to convey prosody scores back to the user to assess a child’s reading ability. Whether a child paused in the right place after a comma or period or raised their intonation at the end of a question are all important indicators of a child’s reading fluency.
4. What are the options for removing or deleting data after an API call?
Our API services offer the option to “delete immediately after processing,” meaning voice data won’t be stored in our systems.
Also important to note is that clients are required to send us anonymised audio, so that any time voice data is stored in our systems it cannot be attributed to a specific child.
As a privacy-first company, we’ve worked for many years with safe harbour agency PRIVO, the leading experts in online privacy for minors. PRIVO certifies our systems and procedures regularly to ensure they continue to protect kids’ fundamental right to data privacy.
5. There are a lot of toys and voice-activated products that are not prioritizing privacy, especially for children. What is SoapBox’s vision for how voice technology could be used ethically in consumer-facing and play-based products?
As toy and games companies become more interested in voice-enabling their products, they’re also becoming more educated about privacy issues.
Kids’ voice data sent using the SoapBox API is never re-used, shared, sold, or stored for marketing or advertising purposes. Our technology is also available as an on-device solution — think robot voice-controlled via an app on your phone or tablet — or as an embedded solution on a chip inside a teddy bear a younger child wants to cuddle and talk to.
For more on our vision for ethical and private voice tech for kids, read our Five Calls to Action for the Voice Industry.