Improving Literacy One App at a Time | Keen Research Blog

Improving Literacy One Step App at a Time

How Educational Technology Companies Embraced Keen Research Offline Speech Recognition Solution

Author: Ognjen Todic | September 28, 2022

The value of literacy is far-reaching – for individuals and society. Literacy is good for your health, builds skills, improves the economy, promotes gender equality, strengthens democracy, and enhances quality of life.

And yet, for all its benefits, improving literacy is still a challenge. According to the U.S. National Center for Education Statistics, 34% of students score below reading level in fourth grade. And while most states have seen childhood literacy improve since 2003, five states have seen it decline.

It’s not just a problem for children, either. The U.S. Department of Education reports that 54% of U.S. adults – 130 million people – lack proficiency in reading. And a 2020 study by Gallup revealed that low levels of adult literacy cost the country as much as $2.2 trillion per year.

What’s to be done about it? Parents and teachers play the most important role. For example, in my own family, my wife (a teacher, not coincidentally) and I began reading to our two daughters when they were still infants. Those early seeds grew into a full-blown love of reading that continues today, a decade later.

But not all families understand how vital literacy is. Socio-economic status can have a negative impact, and with adult literacy a problem, some parents are not well-equipped to instill that love of reading. At schools, a lot of teachers are already overwhelmed and stretched thin, especially in disadvantaged communities. The fact is, we need to provide all kids with access to books, make it easy for them to practice their reading skills, and to develop the joy of reading and exploration.

That’s a belief we at Keen Research share with LeVar Burton, literacy activist, long-time Reading Rainbow TV host, and newly-named Chief Reading Officer of Osmo. Osmo is using Keen Research speech recognition technology in its just-launched Reading Adventure product to make learning to read more fun and exciting for children. Part of global EdTech company BYJU’s, Osmo products are used in more than 2.5 million homes and 50,000 classrooms.

EdTech Embraces Offline Speech Recognition

Reading Adventure is just the latest example of how the Educational Technology market has embraced the use of KeenASR SDK for on-device speech recognition. That’s not surprising. Before starting with on-device speech recognition at Keen Research several years ago, I spent a number of years at Ordinate Corporation where we developed the first automated spoken language test – and speech recognition was a core part of the platform.

I was well aware of how speech recognition technology could be leveraged for different EdTech uses. What I didn’t realize was how speech recognition could engage new readers by helping to create magical experiences for them.

That realization came several years ago when we started working with Novel Effect. They had already built their augmented audio reality app for storytelling, but needed a reliable and accurate on-device speech recognition engine; that’s where Keen Research jumped in. The result was purely magical. While a parent (or teacher or student) reads a real book, the Novel Effect app tracks where they are in the book – using KeenASR SDK – and plays corresponding audio effects or “soundscapes,” as the company calls them. This creates an immersive experience and pulls kids into the story.

For example, as you read “…and the lion roared,” kids hear the sound of a lion’s roar from the speaker on the smartphone. What makes this really magical is that the technology is completely hidden from the user – the focus is on the book and reading. The technology augments and enhances the experience by engaging more of the senses.

The voice-interactive EdTech magic continued with Rally Reader. This digital reading coach offers a plethora of digital books and motivates kids to spend more time engaged in reading; it’s been scientifically proven that 15 minutes or more of engaged reading per day delivers significantly higher literacy scores.

Next was Readability Tutor, which provides a reading and comprehension learning mobile app, focusing on young and struggling readers. Their app is focused on improving reading skills and fostering a lifelong love of reading.

The common thread for almost all of these mobile apps is that they allow kids to practice reading on their own and get real-time feedback. Using the KeenASR SDK for on-device speech recognition enables these apps to know when and what the kids are reading, and how well they are reading.

In between working with these customers we’ve helped other EdTech companies voice-enable their mobile apps focused on children. Nickelodeon uses KeenASR SDK to power interactive videos in their Noggin app; kids can interact with characters in videos using their voices – and learn in the process. PBS Kids built a STEM-focused app called The Cat in the Hat Invents that uses KeenASR SDK for voice control of a little robot that moves in the game. Almedia developed a mobile app focused on promoting cross-border cultural collaboration between Germany and France; their app helps students practice speaking German and French.

Can Speech Recognition Help Children Learn How to Read?

When we started signing up our first EdTech customers, our SDK features were not as rich as they are today, yet the SDK was still useful. For oral reading tasks, even basic information like a solid estimate of reading accuracy and the rate of speech provides tremendous insight into a child’s reading ability.

Since these early days, we have added many exciting features that enable our customers to develop better products and help kids learn how to read. Some of these features include:

  • Acoustic models optimized for kids’ voices
  • The ability to recognize incomplete words
  • The ability to model common mispronunciations for words
  • Goodness of pronunciation scoring

In addition to these features specific to oral reading instruction and assessment and language learning, the fact that all the processing occurs locally on the device means that products using KeenASR SDK can have:

  • Guaranteed COPPA compliance
  • Low latency
  • Scalability
  • The ability to customize the SDK to specific use cases

Finally, developers can build their apps in native development environments, like ObjC/Swift on iOS, Java/Kotlin on Android, or use our Unity plugin for on-device speech recognition for cross-platform development.

We are continuously working on improving the KeenASR SDK, from core speech recognition performance (accuracy and speed), to providing additional features that enable our customers to offer better quality apps to their users.

It’s gratifying to see how products developed with our technology have been embraced by children, parents, and teachers. Our future depends on their success and we are proud to support them. Above all, we are proud of our customers and how they are using the KeenASR SDK to help raise literacy levels across the United States – and beyond.

We provide professional services for SDK integration, proof-of-concept development, customization of language and acoustic models, and porting to custom hardware platforms.