In this project we aim to classify toxic texts and try to detirmine which text is toxic.
With an increase in the need of online communication, multiple websites and applications have integrated an in-application messaging option. It is a very convenient and useful feature but it is also important to discourage any toxicity. Our aim is to make a toxicity detection module which can predict the amount of toxicity in a particular text. This module can help identify any toxic texts or even people which can then be reported to the authorities.
Our objective is to create an application which can detect toxic text using Sentiment analysis techniques namely Word embedding based classification. We will use GloVe Vector embeddings to embed the words and then use ml models to classify the text. We aim to train the model using the kaggle jigsaw toxic comment classification dataset. This can help us get different values for each type of toxicity providing a more detailed view of the text provided to us by the user. This application will help business to filter toxic text form any portal such as product review pages or other portals where text can be entered and be displayed publicly.
To demonstrate the use of this app we have created a app which can be used in tandom with this app to demonstrate its capability.