Bookbot: A gamified app that provides real‐time, on-device speech recognition technology in Bahasa Indonesian
The UNICEF Venture Fund is featuring members of its A.I (Artificial Intelligence) & Data Science Cohort. In this interview with Bookbot Founder Adrian Dewitts, learn more about Bookbot Technology, which is developing a gamified app leveraging real-time, on-device speech recognition technology that listens to the user reading aloud, providing feedback through pronunciation modeling. With support from the Fund, the Bookbot team is training a speech recognition model specifically designed for Indonesian children.
Bookbot was founded by Adrian DeWitts after he discovered that his son, Forester, was falling behind in primary school with his reading and writing skills. Adrian built an app to listen to Forester read aloud and guide him by highlighting incorrect words. When Forester graduated from primary school, he won a reading award. Following this initial success, Adrian launched Bookbot to help other children learn to read.
Last year, Bookbot was invited by INOVASI, an Australian government aid program, to bring Bookbot to Indonesia. The Bookbot team have now created over one thousand phonics-leveled books (which accelerate reading skill) for Indonesian children. These books are based on the interests of users, using a speech recognition program that listens to these children as they read aloud. The technology works on mobile phones and can be downloaded and used offline, so that children can access books even when they don’t have internet connection.
How would you describe your solution to a non-technical person?
These days, reading, writing, and other communication skills are essential for leading a happy and successful life. Many children find it difficult to access books in rural Indonesia. To address this problem, we designed Bookbot to work on devices that were common in Indonesia, such as mobile phones. We also made sure that Bookbot could run without internet connection, further increasing accessibility.
The Bookbot app listens to children read aloud, tracking their reading fluency (the ability to read with speed and accuracy) and the time they spend reading. By tracking reading fluency, teachers can work out how much support a child needs. This information also helps teachers decide when to increase the challenge by progressing to more difficult books. In addition, tracking allows us to provide incentives for children, like certificates. However, our main goal is to create intrinsic motivation for reading: that is, learning to read for the satisfaction of reading itself, not for outside rewards.
"The Bookbot app is helping children with limited access to books across Indonesia learn how to read. Our speech recognition technology has also shown incredible results increasing the reading fluency of children with learning difficulties."
What type of social issue/ impact area does your solution aim to solve?
The World Literacy Foundation says that over half of all children truly struggle to read; either through learning difficulties or a lack of support from a competent adult. As a result, today 1 in 5 adults globally still cannot read or write.
Globally, 584 million children are experiencing reading difficulties and more than 393 million children have failed to gain the basic literacy skills at age 10. At Bookbot, we are helping children learn to read, especially those without adequate reading infrastructure (physical libraries, books, etc.) and helping children with reading difficulties, such as dyslexia, learn to read.
In the past 18 months, we have been working closely with the Indonesian Ministry of Education. In a 2016 survey by the Central Connecticut State University, Indonesia was ranked 60th out of 61 participating countries for reading comprehension. Previous studies and discussions show that several of the reasons behind this low ranking include limited accessibility to books, along with unattractive, overly formal, and lecture-like book topics.
What's promising about the use of artificial intelligence and data science technology?
Currently, speech recognition technology is often targeted at adults and is usually online. In the case of Indonesia, this technology has a high error rate. Having a speech recognition that caters for children, offline, and with a low error rate can help young Indonesians practise their reading. This technology helps with reading self-sufficiency, making it useful in scenarios where parents cannot help their children or teachers need to support many students. We can also use children’s speech recognition for other early learners who do not know how to read. We have made this technology open source in the hope that others will take advantage of it and create additional educational literacy products for children in Indonesia.
What is unique about your solution and how is it different from what currently exists?
Bookbot differs from other reading apps by helping children keep their focus on the words they are reading, improving their pacing and their comprehension. The app listens to the child as they read aloud, assisting their reading, and tracking their fluency. Through the app, we also provide access to a large library of diverse levelled phonics books in a small space, available offline. In addition, our reports app allows teachers and parents to see reading fluency improvements over time.
Why does being Open Source make your solution better?
Bookbot is committed to accelerating the attainment of the Sustainable Development Goals in Indonesia. As a result, we have made the new technology we have developed in Indonesia, such as our speech recognition program, available to other developers to use or improve through open source. By helping other organizations that are working towards our common goal of literacy development, we can generate impact that goes beyond the children using our app alone. This also facilitates collaborative learning and promotes organizations working together in alignment with SDG (Sustainable Development Goal) 17 - Partnerships for the Goals. At Bookbot, we recognise that everyone needs to work together to achieve our collective goals.
Tell us more about your team. What makes your team diverse?
We currently have a team of 16, brought together by our passion for literacy and giving all children the opportunity to learn to read. Our team members are located across Asia, from Vietnam and Indonesia to Singapore and Australia. Each member of our team is at a different stage of their life, with some as young as 20 and still at university, while others have 20 or more years of experience in the industry. While we come from vastly different backgrounds and upbringings, we are united in our commitment to the Bookbot cause.
Why is diversity important for your startup? How does it add value?
Diversity is integral to every startup and Bookbot is no exception. Team members from different backgrounds bring with them their own unique skill sets and knowledge. It’s important to recognize that everyone brings different strengths to the table. These different perspectives promote increased creativity, innovation, problem-solving, and decision making. In a broader business sense, the benefits include aspects such as intercultural fluency and knowledge of international markets, which are invaluable for startups looking to expand globally.
What do you plan to do with UNICEF's Venture Fund investment and how will you use that to leverage raising follow-on investment?
We are using the funding to create an automated voice data collection and organization pipeline. This will remove adult voices, and splits information into smaller files that can be used in the speech recognition model training system. Our team is training a speech recognition model specifically designed for Indonesian children. We’re also running an efficacy study to monitor improvements in children’s reading and literacy skills using Bookbot. Bookbot will make the speech recognition model and open-source code available to others, so that they can take advantage of the technology. We’re building Bookbot for Windows laptops and Chromebooks, which are also common in Indonesia. In addition, we’re running an integration study with the Ministry of Education to see how Bookbot can work best in classrooms. Our work is funded by the UNICEF Venture Fund and INOVASI, and we are seeking more support from the Ministry to continue our programmes in Indonesia. In this period, proving the efficacy of Bookbot will help us raise further investment.
"The support from the UNICEF Venture Fund is helping us prove the efficacy of Bookbot as a learning assistance tool, in turn helping us to raise further investment from other sources."
What challenges are you currently facing in building your solution and/or startup?
Data collection: we need large quantities of data to build a great speech recognition model. While the amount of data we have collected so far is wonderful, we need a higher magnitude to achieve a speech recognition system with higher accuracy.
How can others support you in working towards overcoming these challenges?
If more children use Bookbot, it will help us obtain more data. We’re looking to incorporate more regional accents across Indonesia, to help children from across this diverse country learn to read.
Bookbot company profile here.