AWS Public Sector Blog
How artificial intelligence and Amazon Alexa are teaching students to write
The three Rs—reading, ‘riting, and ‘rithmetic (reading, writing, and arithmetic)—first appeared in print in The Lady’s Magazine in 1818. While the world has changed in 200 years, the three Rs still hold true as critical academic, career, and life skills. At Amazon, writing is an integral part of our company’s culture and is the primary mechanism used when making decisions. One thing that has changed in 200 years is now machine learning (ML), artificial intelligence (AI), and voice technology can help improve student writing.
Ecree, an AWS EdStart Member, was founded by educators who believed the way we teach students to write is broken. The volume of work a teacher is being asked to do means students don’t always get the necessary attention and feedback required to develop their writing skills. Ecree’s founding team set out to support teachers and students by creating the world’s first rules-based tool that assesses student writing in the same way a highly trained professor does, on-demand. To date, Ecree has supported nearly a million submissions.
“Students don’t get enough useful feedback. Ecree solves this problem: it trains students to write by giving them great feedback almost instantaneously on as much writing as they want to submit. And it frees up teachers to do the teaching they all love. Education needs Ecree now more than ever,” said Ram Neta, professor of philosophy at The University of North Carolina-Chapel Hill, and Ecree user.
Migrating to improve reliability
Ecree migrated to Amazon Web Services (AWS) in 2018 from another cloud provider after experiencing database outages lasting over eight hours and waiting days wait time for technical support. Ecree moved to AWS for its reliability, depth of offerings, and customer-centric approach.
Since migrating, Ecree improved the speed they can deliver feedback to students by 90 percent while reducing processing load by 90 percent and reducing costs per student by more than half. These efficiencies were realized through Ecree’s use of Amazon Elasticache for Redis, which brought latency to virtually zero, and relational database Amazon Aurora.
Utilizing machine learning for reading feedback
Ecree’s rules-based artificial intelligence engine assesses a student’s essay based on pre-defined rules. The rules Ecree uses are the standard elements of writing rubrics, which Ecree publishes so that students and teachers know exactly what the platform measures. Feedback is based on these rules and is driven by what the student has written. Feedback is delivered at the specific textual point where it is relevant. Ecree uses AWS Lambda, which runs code serverless, to trigger this rules-based feedback for students.
Ecree uses ML encoding and decoding neural network models to decode a sentence to derive its meaning and then encodes the sentence into a simpler form. The ability to be use Amazon Elastic Compute Cloud (Amazon EC2) P2 Instances to spin up and spin down graphics processing units (GPUs) as needed is critical to effectively run these and other ML models.
“AWS offers a product for almost every conceivable need we could have. For any scenario we have faced, there are a million solutions explained online. As a growing company with varying concerns, sometimes we need to do things quickly and pick up the pieces later, and other times we need to do things more elegantly and move slower, but no matter what, AWS always helps. I don’t have to worry. I know I have a team of people at AWS ready to respond if I need anything,” said Robin Donaldson, co-founder and chief technology officer of Ecree.
Going serverless to ease deployments and scale
Ecree uses Amazon Elastic Container Service (Amazon ECS) with AWS Fargate to run and scale code without having to worry about configuring servers or compatibility with the underlying software. Ecree has been coding complex AI and ML models and treating them as services, decoupled from the underlying hardware and software. Ecree is migrating from their original monolithic Personal Home Page (PHP) codebase to AWS’s lightweight microservices-based architecture to ease deployments, scale services independently, and increase agility. “Using Amazon ECS to containerize has made things much finer grained,” according to Robin, “In terms of scalability this is gold.”
Asking Alexa to improve student writing
During re:Invent 2019, AWS’s annual learning conference for the global cloud computing community, Ecree’s ceo and founder Jamey Heit sat down with AWS EdStart leadership to discuss business needs. Through this executive conversation the Ecree team began exploring the power of voice and was connected with the Alexa in Education team at AWS.
“AWS EdStart has some of the sharpest minds in education supporting its member companies. The team combines the reach of a large organization with the thoughtfulness, grit, and creativity of small companies. Their innovative, rigorous, and thoughtful approach makes a real difference in addressing the challenges and opportunities in education,” said Jamey Heit, co-founder and chief executive officer, Ecree.
During the ideation phase of building an Alexa skill, the team put themselves in the shoes of a writer. Writers block and second guessing how a sentence sounds were consistent challenges. The Ecree team wondered—could they ask Amazon Alexa to help you write the sentence properly?
Voice technology is a natural extension of the Ecree platform. They will be launching an Alexa skill to provide students with a new modality for access and feedback to improve their writing. Students will be able to ask Alexa to review their spelling, grammar, and style. In addition, the Alexa skill will provide suggested simplifications to complex sentences, and even supply quotes and citations from literature and historical text.
Learn more about AWS EdStart, the AWS educational technology (EdTech) startup accelerator, designed to help edtechs like Ecree.
Want to know more about how EdTechs are using Alexa to deepen engagement, improve student learning, and make life easier for teachers and administrators? Visit Alexa in Education.
And learn more about AI and ML for education.