Abusive-Language-Detection

Web App

Web App is accessible through this link: Abusive Language Detection

Description

This project is made for the detection of the Abusive Language from text provided by users. The model is trained with latest Sklearn Version 0.24.1 on the Abusive/Non Abusive dataset made by the combination of various publicly available datasets.

Classifiers

For this project we trained 3 models:

1. Random Forest
2. Calibrated Classifier
3. Decision Tree

Calibrated Classifier performed well as compared to other two classifiers. For final model, Calibrated Classifier was trained.

Censor Abusive Words

To censor abusive words we have used the Spacy PhraseMatcher which matches words of user input to the bad_words.txt and replace the matched words with asterisk(*).

Flask App

For deployment of the website we have created an HTML website with . The model works in real time, which means for every typed letter it will make prediction for the sentiment of the whole sentence.

Abusive Demo

Censored Demo