AXON
The Learning App
Development
2nd April 2024 - Present
(chapter 1) : Why
A Few Days Before we made this, a notice came up in our college group, it was an announcement for the Hackathon "Spectrathon",it was a Hackathon Held by the Goa College of Engineering. i saw it and i realized that id never really taken part for an Offline Hackathon , id lead a team to organize one, but id never participated myself, sooo , i did and i formed a team with a few of my friends and we went thru the problem statements.
Choices?
The Problem Statements consisted of around 20 statements which were either about AI, CNNs,LLMs and some others. this was the time AI and LLMs were Famous, people knew more about AI than when their dad would get back with the milk "other words".
so we had to choose, the work with CNNs seemed very interesting, the statement was to analyze camera footage to find out crime-related information, you know the Detective kinda shit, and i don't. know bout you, but i feel everyone likes (the thought of) working with CNNs and Images, However, although i have a moderately good understanding of it , i decided against it. luckily my team mates agreed. and we settled on...
Artificial Intelligence
Domain specific Learner facilitator Assist Software Given a learning resource the application should assist the learner in preparing for exams. For the facilitator the teaching material should be prepared automatically. The learning resource could be electronic materials on a particular subject.
(chapter 2) : How
Choosing is Easy, But Knowing How to do it, is easy if you're a nerd - Someone smart
1) Data Collection
So here's where we did 60% of the work. our first goal was to do a lot of Research, so we got a shared doc, and started pasting anything that seemed relevant, wether it was a link from hugging face or some algorithms from sentence transformers or something from ChatGPT.
2) Testing & Planning
Next we needed to check out what works and what doesn't, we had amazing ideas of what could be done,but realistically we needed to remove stuff and finalize that, here @ryzxxn figured out the llm with mistral was okay at best, and we couldn't rely on it to give us what we needed every time, so i came up with an nlp pipeline, where we extracted the parts of speech using the POS in NLTK Library and convert it to our desired form the implementation of this was done by @justSammy1604. Another feature we wanted was to find a video explaining the topic, this was done by @Terrence31
Work in Progress ...