Tat | Devpost

Inspiration

We got the inspiration while solving some math questions. We were solving some of the questions wrong, but couldn't get any idea in what step we were doing wrong. Online, it was even worse: there were only videos, and you had to figure all of the rest out by yourself. The only way to see exactly where you did a mistake was to have a teacher with you. How crazy! Then, we said, technology could help us solve this, and it could even enable us to build a platform that can intelligently give the most efficient route of learning to each person, so no time would be wasted solving the same things again and again!

What it does

The app provides you with some questions (currently math) and a drawing area to solve the question. While you are solving, the app can compare your handwritten solution steps with the correct ones and tell if your step was correct or false. Even more, since it also has educational content built-in, it can track and show you more of the questions that you did incorrectly, and even questions including steps you did incorrect while solving other questions.

How we built it

We built the recognition part using the MyScript math handwriting recognition API, and all the tracking, statistics and other stuff using Swift, UIKit and AVFoundation.

Challenges we ran into

We ran into lots of challenges while building all the data models, since each one is interconnected with the others, and all the steps, questions, tags, etc. make up quite a large variety of data. With the said variety of data, also came a torrent of user interface bugs, and it took some perseverance to solve them all as quickly as possible. Also, probably the one of the biggest challenges we dealt with was to deal with the IDE itself crashing :)

Accomplishments that we're proud of

We are proud of the data collection and recommendation system that we built from the ground up (entirely in Swift!), and the UI that we built, since even though the app doesn't have a large quantity of educational content inside yet, we built it with the ability to expand easily as content gets added, in mind.

What we learned

The biggest thing we learned was how to build a data set large enough to give personalized recommendations, and also how to divide and conquer it before it gets too complex. We also learned to go beyond what the documentation on the internet offers while debugging, and to solve things by going from examples, without documentation on how to implement.

What's next for Tat

We think that Tat has quite a potential to redefine education for years to come if we can build more upon it, with more content, more data and even the possibility of integrating crowd-trained AI.