search

b4apple / Nicum-Question-Tool

Named after the honorary Nicum sir, this tool allows you to have a Question Attempting UI for question papers whose PDFs or Printouts you may have
Other
1 stars 1 forks source link

Discussion #1

Open Kishlay-notabot opened 4 months ago

Kishlay-notabot commented 4 months ago

Hey there, Kishlay this side, I want to know that what kind of features do you want to add to this webpage? And what's the roadmap. I can help you with the development, if the domain matches my skillset.

Kishlay-notabot commented 4 months ago

Edit: I think the pdf segmentation could be done without AI, but it needs to be specific. Will discuss more soon.

b4apple commented 4 months ago

Hi @Kishlay-notabot thanks for your interest.

  1. Immediate Features : Matrix Match 2 x 2 UI element and subsequent response display. 2. 50% of the page on the left to show the Question UI, 50% on the right to display the pdf in a pdf viewer element.

If you can implement these two features, extremely helpful! The roadmap beyond this is in the making and I shall get back to you in 2-3 days.

Kishlay-notabot commented 4 months ago

I think, running OCR on the pdfs imported, and then seperating each question with its own set of MCQ radio options as A,B,C,D and making them be in a "gallery" just like NTA exams where you click next to get to the next question. I can work with the OCR part. But the algorithm which will seperate the questions maybe can be based on the numerical numbering? I haven't brainstormed much into that side. But I think using OCR can help. Doing this would be better suiting for students than a 50/50 content and pdf page. Because then you manually need to scroll both the sides, MCQs and the Pdf too. Let's see how this goes, I'm just seeding an idea to you.

b4apple commented 4 months ago

sounds good, can you prototype the OCR part?

Kishlay-notabot commented 4 months ago

Running OCR on imported pdf is dead easy.. I've worked a lot with Tesseract engine and it can be done easily even in the browser. But working on the algorithm where we seperate each an every question and assign it to html elements should be tricky. That part is very specific to the type of pdfs kids generally use. You should get some samples of what sources do they take it from, and analyze the physical structure of how the elements are placed. It would be hard to generalize an algorithm for variety of pdfs.