Ryan and Elias - Le Filter (Hate Speech)

Inspiration

We were inspired to create this project from hearing about how some apps use machine learning to detect spam emails. This inspired us to make something similar for detecting hate speech.

What it does

This app is a very simple machine learning model that can scan a site for content that can be considered hate speech. It can be configured to flag different words, and trained using different input data.

How we built it

We built this app using Python, mainly utilizing scikitlearn and numpy.

Challenges we ran into

Originally, we were going to use Java or Rust, but we decided not to use Rust because we weren't both familiar with it, and switched from Java to Python because Python provides a much more complete ecosystem for machine learning and a much simpler syntax.

Accomplishments that we're proud of

We're proud of how we were able to deploy a basic machine learning model and use it in a semi-practical setting.

What we learned

Prior to this, I had never worked with creating machine learning models from scratch. This was a new experience for me, and learned a lot about it while completing this project.

What's next for le filter

If we had more time, I would like to do two things:

1. Rewrite at least part of the app the app in Rust. This would provide better performance, and I generally like Rust more to begin with, although developing with Rust usually takes more time, especially since neither me nor my teammate are that experienced with it.
2. Replace our current linear regression model with a classification model to get more accurate and nuanced results. We did what we could in the time constraints, but in the future I think training a proper model would hugely benefit this project.

Built With

• numpy
• python
• scikit-learn