-
-
How It Works
Inspiration
Over the past few years, the art of learning has been slowly reduced to a chore rather than something people look forward to. Perhaps it was a factor of the pandemic, or maybe it is indicative of a dwindling education system. However, we believe that learning should be an exciting journey and have a simple to follow path.
What it does
As a result, we created TagTeam, a web application that uses natural language processing and optical character recognition libraries to tag users’ notes with whatever they write about, whether it be math, physics, or even history and then suggest further relevant readings.
How we built it
The program pipeline works as follows: Our flask backend stores and retrieves documents from a MongoDB atlas cluster to perform operations such as JWT authentication and object relational modelling We interact with the google cloud vision and pipe the data from it to the Cohere and OpenAI APIs to do some cool NLP stuff Then our React frontend simply fetches and posts data accordingly using Axios
Going in more depth about the APIs we used: The google cloud vision API used confidence level based OCR to parse a base64 encoded image into a simple string That string is then corrected using Cohere’s text correction model, and is then passed through two more xlarge model layers to classify and generate tags for the image Finally, the corrected string is also asynchronously piped to the openAI davinci-003 model to give recommendation based on the content.
Challenges we ran into
For majority of the team, this was our first hackathon and for some was even our first time trying web development. Resultantly, coordinating and delegating tasks was hard due to everyone's difference pace of completion. Also, Google Cloud authentication is a nightmare. Also, BSON data is a nightmare to work with. Also staying up until the next day was a nightmare.
Accomplishments that we're proud of
"Completing" the project and overcoming whatever obstacles stood in our way.
What we learned
A bit of everything, from how to design a REST API to how to manage stateful applications.
What's next for TagTeam
We are planning on further polishing this app to be suitable for mass use and adding features we unfortunately did not have time to add, such as recommending courses/books and a friend system.
Log in or sign up for Devpost to join the conversation.