Dentsu and AWS: “Easy Reading” books for developmentally disabled (IDE102)
Easy Reading: Unlocking Access to Literature for People with Learning Disabilities
Introduction
Jeremy Bević, Solutions Architect at AWS, and Thiago Winkler, Head of Innovation at denu creative Brazil, collaborated on the Easy Reading project to make literature more accessible for people with learning disabilities.
The presentation covers the journey of building this solution, the key learnings, and the future plans for the platform.
About denu
denu is a 123-year-old Japanese group with agencies and operations across different countries.
The Innovation team at denu focuses on exploring new technologies, including AR, VR, and AI, to create innovative solutions for their clients.
The Easy Reading Project
The project was initiated by Visit Portugal, who wanted to explore new ways of making classic books accessible to people with intellectual disabilities.
Approximately 200 million people worldwide experience difficulties when reading literature, and the existing "Easy to Read" standard is a lengthy and expensive process.
The goal of the Easy Reading project was to develop a solution that could convert books into an accessible format using AI in a matter of hours, instead of weeks.
How it Works
The Easy Reading platform takes the original book content, simplifies the text, generates relevant images, and formats the content to meet the "Easy to Read" standards.
The process involves using various AI models from Amazon Bedrock, such as Anthropic's Claude family of models for text processing and generation, and Stable Diffusion and Titan Image Generator for image creation.
The output is reviewed by an "Easy to Read" specialist to ensure the content is suitable for the target audience.
Key Learnings
Generative AI is a fast-moving space, and the team had to adapt to three major model updates within the 3-month project timeline.
Using the right model for the right task was crucial, as different models had varying capabilities and context sizes.
Asking the models to perform tasks beyond just text generation, such as rating content against a set of rules, helped optimize the output.
Collaboration with the rights holder and the "Easy to Read" specialist was essential to ensure the final content was of high quality and suitable for the target audience.
Optimizing for cost and scaling the solution were important considerations, as the goal was to serve a large population of 200 million people.
Next Steps
Explore supporting multiple content formats, such as audiobooks and podcasts, to cater to different accessibility needs.
Utilize the new AWS Bedrock Flows service to improve the orchestration of the content conversion process.
Continue to adapt to the fast-paced changes in the generative AI landscape and leverage the latest model improvements.
Conclusion
The Easy Reading project showcases how AI can be leveraged to make literature more accessible for people with learning disabilities.
The collaboration between AWS and denu demonstrates the potential of AI for Good and the importance of human-AI collaboration in developing such solutions.
The team encourages others to explore similar use cases and challenges the audience to bring their own ideas to life using the available AI technologies.
These cookies are used to collect information about how you interact with this website and allow us to remember you. We use this information to improve and customize your browsing experience, as well as for analytics.
If you decline, your information won’t be tracked when you visit this website. A single cookie will be used in your browser to remember your preference.