ReadingDojo

Inspiration

I was inspired to create this tool by the need to streamline the process of generating custom reading comprehension materials. I myself struggled with reading comprehension when I was young, as English is not my native language, and I was certain that many others have the same problem as well; it could be in a foreign language for an English speaker, such as French or Spanish.

For educators, traditional methods of creating custom multiple-choice and short-answer questions to check for understanding can be time-consuming and labor-intensive. Sure, there are ready-made worksheets, but they have huge drawbacks such as the lack of customizability and sometimes the problem of pay-to-access.

For parents, they mainly assign reading worksheets to their children to help them succeed in school and just build fundamental reading skills in general. To do that, many may just scout a boring, uninteresting resource online - something that children may not find interesting and instead think of as a chore to get over with.

To address these problems I decided to build a web-application that would take a shot a providing an efficient solution for all three audiences -- for students, for educators and for parents.

What it does

My solution, Reading Dojo, harnesses the power of generative AI to write customized questions and passages that check for students' understanding of a particular reading, which can be generated by the tool itself or be sourced from custom text such as an excerpt from a novel. As of now, the tool allows its users print out a worksheet and answer the questions online for immediate and convenient scoring.

Use Cases

For students: Reading Dojo can be used to help students improve reading comprehension in different languages. Reading Dojo supports English, French and Spanish. The tool can help students prepare for school where they can practice with real texts from English (or any class in general) class to better understand the text. Students can use the tool to better interact with the text and stay focussed while reading, helping them prepare for an assessment.

For educators: Reading Dojo can help educators generate a set of multiple choice or short answer questions, tailored to match their needs with the help of AI. They can then use the content generated to in multiple different ways: assign reading homework (for which answers can not be searched up online), receive guidance in creating assessments, and more.

For parents: Reading Dojo can help parents create reading worksheets that their children will enjoy doing by generating tailored questions (and passages) from children's favorite books or cartoon characters, depending on the age level.

How I built it

Making Reading Dojo come to life involved a lot of drinking water, a lot eating, a lot of StackOverflow, a lot of slacking off, a lot of frustration, and of course, a lot of coding. Jokes aside, I built it with a lot of passion and dedication.

Here is how I did it:

Design draft of User Interface using Figma.
Create-react-app.
Tackle one feature of the tool at a time (functionality & design).
Constant testing & debugging.
Tutorials + StackOverflow + Music (for entertainment).
Repeated steps 3-5 till I was able to fit in everything I had in mind for the hackathon.
Deployed to Netlify.

Challenges I ran into

The most daunting challenge I came across while building Reading Dojo is that I haven't used OpenAI's APIs before and this hackathon is the first time I'm using the GPT model. Since the soul of the project relies on the GPT model to generate questions and responses, I had to come up with the best possible prompt to provide the best results to the users. It was extremely frustrating, but all of that frustration is part of the learning.

I also ran into an issue that lead to reads and writes count skyrocket in Firebase. In fact, in a matter of few seconds, my read count went up to 48k, just short of Firebase's generous free tier offering of 50k reads per day.

Accomplishments that I'm proud of

Managed to integrate OpenAI's GPT-3.5 model in my application through a backend server using express.js.
Adding multiple functional features to an application in a short time-span as a single developer.
Creating a decent 2 minute product showcase video.
Being able to finish and submit a working project for my first hackathon.

What I learned

I learned to utilize the GPT-3.5 model and Firebase's features more efficiently.
I learned that working solo is not at all easy and takes a lot of time and commitment.

What's next for ReadingDojo

I can see potential in Reading Dojo if I add more features such as explanations for answers, analytics for students, custom grade levels, and more customization for users. Certainly, I'll work on implementing all these features with the help of others because I've fallen in love with creating this tool during this hackathon. I'll also try to improve the results of the actual generation by spending more time tuning the model.