Yik-Yak Alternative

Yik-Yak Alternative allows people to communicate with their peers anonymously, and replaces negative messages with the help of machine learning while maintaining the post's original content.

Comment

Inspiration

We wanted to create a positive alternative to the popular anonymous app yik-yak.

What it does

It uses Large Language Models to perform sentimental analysis on every message and manipulate negative message to be more neutral.

How we built it

  1. Backend: We developed a flask app to handle get and post request for messages. We stored messages as json files due to its convenience considering a hackathon environment.
  2. Machine Learning: We implement two pre-trained models from Hugging face to perform sentiment analysis and style transfer from negative to neutral text. We developed two functions. The first one screens messages and perform sentiment analysis to determine whether the text is positive or negative. If it was negative, we ran it through a second function to modify it to more neutral tone.
  3. Frontend: We designed an React web app to handle submitting messages and displaying them in real time.

Challenges we ran into

Firstly, we had trouble finding a model that is able to convert negative messages to positive or neutral ones. Additionally, we had trouble finding a model that was lightweight enough to run locally. Secondly, we had trouble updating the main page to display the added messages in real time.

Accomplishments that we're proud of

We are proud that we were able to find an appropriate model to convert negative messages to neutral, and that we successfully launched a web application that can receive, edit, and display message in real time.

What we learned

We learned to collaborate over git and utilize virtual environments in this project. We learned the importance of following git conventions while coding by having to work through overwritten code.

What's next for Yik-Yak Alternative

Trying to implement a different or more efficient model to convert text to positive instead of having to stop at neutral. Building a more robust frontend for our web app. Implementing a database through technologies such as MongoDB or SQL to store messages. Develop further features for our users.

Share this project:

Updates