NLP: Multilanguage Toxicity Detection

Build a deep learning model to detect toxicity in sentences in any language and deploy it on ShinyApps
deep-learning
NLP
data-science
Author

Vertikal

Published

November 10, 2023

Embarking on the adventure of natural language processing, my recent project aimed to tackle the challenge of sniffing out toxicity in sentences. Armed with a dataset I snagged from Kaggle, I set out to build a NLP model from scratch using tensorflow & keras. The main goal? Classifying sentences into distinct toxicity categories like TOXIC, SEVERE TOXIC, OBSCENE, INSULT, and IDENTITY HATE.

It’s not just about the code and algorithms – this project is a small step in making the online world a more pleasant place by sifting through and addressing various shades of unfriendly language. I must admit that the model I’ve been using isn’t quite powerful enough due to a shortage of resources, such as the dataset used and the computational power required for training the model.

Click the picture below to run the app.

Your Image Description

For source code : Github

For dataset : Kaggle