“Does your voice technology cater to different accents and dialects?” This is the most common question we get asked by new clients. And no surprise, since all education or entertainment brands worth their salt are committed to offering voice-enabled experiences that work for kids of all backgrounds, geographies, and ethnicities.
The question is well justified however. Last year, a Stanford University study found racial disparities in automated speech recognition (ASR) systems from Amazon, Apple, Google, IBM, and Microsoft. According to the study, these systems misunderstood Black speakers 35% of the time compared with only 19% for White speakers.
SoapBox Labs has set the benchmark for accuracy in voice technology for kids, and that includes catering to globally diverse accents and dialects.
Since our founding in 2013, we’ve worked to build a voice engine that offers 95% accuracy, removes the risk of false positives and negatives, and understands the voice of every kid you’d find in, say, a 3rd grade New York city classroom.
Our VP of Speech Technology, Dr. Amelia Kelly, often explains to customers how our acoustic models are trained on accents and dialects from 192 countries, and can accommodate unique or custom pronunciations for any population:
Our models are sophisticated enough to capture differences in kids’ accents and speaking styles, so they will work for kids wherever they’re from and whatever they’re doing when they use our voice engine. –
Dr. Amelia Kelly
Accommodating the complexity and variability of all kids’ voices is a core commitment of SoapBox Labs. Our kids deserve nothing less.
Learn more below or reach out to us: