Hate Speech Detector

Inspiration

I was inspired by a game of Rocket League where the other team was being very rude to each other. This was something that stuck with me because a lot of the players are children, so it's likely that children also participate in this trash talk/rude comments.

What it does

My model detects if a piece of speech is hateful or not.

How we built it

I researched Natural Language Processing (NLP) models, and found a tutorial that uses Keras. I followed this, and was able to create a model. I then went looking for datasets that would help me, and found the datasets I needed. Then, I worked on cleaning up the data, and ran the model.

Challenges we ran into

The code didn't work on my computer for some reason, but it worked on Kaggle. I ended up just using Kaggle to run the code because I decided that it wasn't worth the time trying to debug this issue.

Accomplishments that we're proud of

90% accuracy on the validation set with no parameter fine-tuning.

What we learned

I learned how to create a NLP model, as well as how to use Kaggle to run my code.

What's next for Hate Speech Detector

This can be integrated with a chrome extension so parents can monitor what their kids are writing. There are some issues with this, because it would essentially have a keylogger, which can be mistaken for a virus pretty easily.

Built With

• kaggle
• keras
• python