Personal experince has shown that not everyone learns in the same way, simple rephrasing, differnt method of presentations and having a connection with the material makes all the differnce. Secondairly, how best can we effectively roll out a solution. Unstable homes make using a computer difficult at best, and in areas where internet is not wideley aviable, it needs to be rebust to remove all inequality. Thus, an inital webapp that allows the user to download materials and be able to personally teach them was created.
What it does
Taking data from past papers, the AI system then generates a model on how questions are phrased. We then allow the model to rephrase the terms to suit the learning style of the user, ensuring quality we keep it in check with the inital learning requirements and allowing it to feed it into the linker module for QA testing. Eventually, we hope to expand this to generate simulations and diagrams, even audio. Though this requires better thinking of the offline requirement.
How we built it
We built it using react, as its easy portabilty into electron frameworks, its responsive nature and rampid scaling capabilities makes it the perfect canidate. We also used tensor flow, ensuring that we can add the AI capabilities which were touted. Lastly, we then used MS paint for the art of the presentation.
Challenges we ran into
pdfs. For some reason, pdfs are not standardised, there were three differnt modules for react all name simliar and building off one another but none were able to phrase it into text. 5 hours spent on that, hence the lack of meaningful code. Vaccination, we had to leave for about one hour for the covid19 vaccine during which the issue of pdfs still hadn't been fully resolved.
Accomplishments that we're proud of
Tensor fow working, surprisingly it worked out of the gate without any issues, and with further time I should have been able to better implement it into the system Dynamic page, it had been a few years since I last used react, as such hooks were added and allowed for a static page to be built quickly and effectively.
What we learned
We learned, pdfs are blacker boxes than how the internet runs, tensorflow is quite easily implimented and that react has gotten better with time. We also learned that twenty four hours insn't a lot of time to crush multiple years of research and developmennt into.
What's next for Edu-Net
Continue to tinker with tensor flow and reading data, building up a simple model on how the data should be outputed for the link module would spead up development further. Also, once built, user interfaces and hosting it on google cloud should be the upscaling this product requires.