##How we built it We started by reading serveral CMU and UMich research papers on classification. We prototyped our website with Figma and gathered the machine learning datasets by scraping Twitter and Tumblr with a python script. Afterward, we used Tensorflow, Python, and other frameworks to build two classification sequential neural networks.
##Challenges we ran into Challenges we ran into in the build include creating accurate model, creating clean and accurate datasets for our models, and connecting the AI API to our web application. Creating a semi accurate model was really difficult because often the cleaning, collection, and development process for AI takes weeks. Furthermore, hosting our api was really difficult because we didn't have access to a credit card and no free endpoint hosting software can have tensor flow dependencies. We stayed up all night to best resolve these issues.
##Accomplishments that we're proud of We are proud that we were able to consume over 50+ pages of research paper and gathered 20k data points from online sources. I think it's a testament to what a focused mindset and extensive sleep deprivation can do. Also developing a discord scrapper using the native AI was very difficult but rewarding. Finally, overcoming all the enormous obstacles with creating the three separate machine learning models and incorporating all of them was very exhausting but humbling.
##What we learned Neural networks require vast amounts of data to avoid over-fitting. Also, we should create local alternatives for working on our projects in case we lose connection. This was a big bottle neck for using http protocols to transfer information from our front end to back end.
##What's next for mindful-ai Mindful AI hopes to create larger, cleaner datasets for better classification. We developed a voice chat transcription software which is in our repository, but, we sadly did not have time to incorporate into our project. Mindful AI hopes to add analysis to voice and video on discord with this unimplemented feature in the future.