Re.Wise

Ever find yourself reading through paragraphs of text only to later find out that you didn't really learn any of it.

Comment

Inspiration

Me and my friends being master's students are assigned a lot of readings. We end up going through our assigned readings multiple times and even though at the time it seems like we remember all the details from the paper we end up forgetting answers to most of the questions on the quiz.

So, as a solution we started quizzing each other while reading but this technique wasn't very successful since we either didn't have similar schedules or some of us prefer reading the whole text before revising and others had different preferences.

To sum up, it would be great to have something that could help us revise our readings but with minimal time investment and easy accessibility on the go.

What it does

Re.Wise is a personal friend that works anywhere, anytime and helps you make sure that you learned your assigned readings to score a perfect score on your tests! It's pretty simple to use too because all you have to do is that you need to click a picture of the text you're trying to learn or submit a pdf of the papers and Re.Wise will take care of everything! It will give you options of answering blanks vs answering multiple choice questions! Isn't that great?

Surprise, surprise! We also have a google assistant action that you can use on the go to answer some questions and reinforce your learning! You can also share these questions with your friends to help them also Re.Wise on their own schedule!!

How I built it

First of all I created an android app that allows you to capture photos of the text you want to Re.Wise. This app uses the Firebase cloud ML kit to detect the text in the photo and sends this detected text off to the Django server for processing. The Django server uses nltk toolkit to apply nlp on the text and figure out the context. It then converts text to questions by identifying and replacing the keyword with a blank (_______) or coming up with a question and 4 options for it as a MCQ. The MCQ bit was a little hard since there might be multiple relevant contexts in the sentence that might be relevant as a question and turning it to a question that makes sense is something which people are still working on.

The user can also provide feedback for the quality of the questions which will help improve the NLP engine. The engine is also using machine learning to optimize it's context decipherment and turning text into questions. This is especially required because there is no datasets that work around converting text to questions and so the algorithm isn't very optimized right now for tougher queries.

Challenges I ran into

First of all I was working on a totally unrelated hack till Saturday evening but the idea kind of didn't sound as amazing as it sounded before starting so I had to think of a new idea and implement it within 1 day!

Apart from this logistics issue, I had no idea about the nlp algorithms that would convert text to questions and turns out this is not a very explored area so had to dig into some research papers and get my hands dirty to code those up.

I also had no ideas about how the google cloud platform worked with the new MLkit and the vision api's so had to explore those out but pretty impressed with the results that I got!

Accomplishments that I'm proud of

Got a working product in a day which is actually solving a very real problem in education that every student goes through. Also learned a lot of different stacks and API's while working on the problem like MLkit.

What I learned

Learned Algorithms to convert text to questions, firebase MLkit and cloud vision api's as well.

What's next for Re.Wise

I feel like Machine learning and NLP engines for Re.Wise can be optimized further to spew better results and then there's the obvious improvement on the UI because I really didn't get a lot of time to work on the UI of the app. It's a very minimalistic UI though!

Share this project:

Updates