Toxic-Text-Detection

Toxic texts and Abuses can effect a person mentally. In order to tackle that I have developed this project.

This repository consist of a web application which can classify between toxic and non toxic texts. This uses a deep learning model to classify the text.

Demo: [Click on the image]

The dataset was acquired from Kaggle's Toxic Comment Classification Challenge.
Link: https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge/data

It uses deep learning LSTM model for classification of the text. For a functioning website with deep learning model, I have prefered to use Django Frame Work.

This LSTM model has a accuracy of 96% on test data.

Repository Details:

• The Machine Learning folder contains all the operations performed to build a model. They also contain the trained model saved in '.h5' format.
• The other folder contains Django Files including a fully functioning website in order to test the model.

Potential:

• This can further developed into a API where every one can take advantage of it to avoid toxic texts. By doing this one can integrate into chatting applications, games, ...etc.

[!] Note: API Service has been implemented and this can be used as a API with POST. This will return a Json format of the prediction of a given text.

Backdrops:

• This deep learning model is actually a multilabel classification and has the potential to detect various other labels such as insult,obscene, threat, etc. But however during my testing and according to the data visualisation there is a slight bias towards certain labels. Thats why I prefered to keep it as a binary classifier.

[!]Note: I have generalized all the other labels as toxic and did not skip them. Skipping them would defeat the purpose of the model.