Scholarly Innovation Compilation

Problem Statement (s):

You should basically choose and work on only 1 type of generation only for the assessment, whichever you are comfortable with w.r.t tech stack (Computer Vision, Speech or NLP)

Description:

We are basically looking to build a tool that can generate a lot of new age alt-form content from a research paper as input. You can either build 1 out of the 3 mentioned but build it with as many nuances and user level input customizations as you can think about.

Input:

A Research Paper/ multiple papers

Output(s) Any One of the Output is acceptable:

• PPT (Presentation)
• Formal ones (like conference presentation)
• Explainer fun ones (for wider audience)

Podcast:

• There should be a conversation between two bots such that it should be an easy explanation of the paper in the form of sound.
• Example – NotebookLM

Graphical Abstract:

• https://mindthegraph.com/blog/what-is-graphical-abstract/
• https://mindthegraph.com/create-graphical-abstract/

Video:

• • Reel like learning (less than a minute)– GenZ trending
  • Explainer Youtube video (less than 5 minutes) References like –

• • • Journalclub.io (ref video)

• • • 2 minute papers (link)
    • 3Blue1Brown like with math animation

Deliverable:

API demo with input and output files or workbench like UI where you can input the research paper(s) and get output as any of the variants.

References:

Google Illuminate: https://www.youtube.com/watch?v=59bU5zrgPkc

There is an AI for That: https://theresanaiforthat.com/s/graphical+abstract+generator/

Video Generator overview: https://www.synthesia.io/post/best-ai-video-generators

https://paperswithcode.com/task/text-to-video-generation

Evaluation Details:

• Evaluation will be done based on
• How aesthetically and logically correct any form of alt media is generate (30%)
• Working demo (30%)
• Presentation and Engagement with Mentor(s) (20%)
• Tech Scalability (20%)