Detection of Radicalization through tweets using Machine Learning

With the advent of technology and its continuous growth, social media platforms such as Twitter, Facebook and Instagram can be utilised by extremist outfits to propagate messages and ideologies far and wide in a rapid and efficient manner. The most preferred medium of communication for exchanging id

Project Title

Detection of Radicalization through tweets using Machine Learning

Project Area of Specialization

Artificial Intelligence

Project Summary

With the advent of technology and its continuous growth, social media platforms such as Twitter, Facebook and Instagram can be utilised by extremist outfits to propagate messages and ideologies far and wide in a rapid and efficient manner. The most preferred medium of communication for exchanging ideas and sharing opinions is Twitter - thereby enabling vulnerable individuals to fall prey to the radicalization process - by making the process of reaching out to extremists relatively easier. Users expressing positive ideas about extremism, propagating hate-speech towards entities or expressing allegiance to an extremist school of thought are identified to be at risk. In this project, we will address the need to detect and classify radicalisation risks by employing machine learning models and performing analysis on written text in tweets using traditional Natural Language Processing based approaches.Social Media serves as a powerful tool for extremists to disseminate their propaganda, recruit assets and conduct psychological warfare. Social Media enables religious extremists to exploit the vulnerability factors linked to socio-economic and demographic conditions of individuals thereby making them suitable targets for their radicalization. To curb the rise of radicalism-led militancy using online tools, we use sentiment analysis techniques for the detection and classification of tweets exhibiting extremist tendencies

Project Objectives

1. Identification of tweets embracing Islamic Fundamentalism.

2. Measuring the performance of Machine Learning models such as Support Vector Machine, Random Forest, and Deep Neural Networks in identifying radicalized tweets.

3. Implementing State of of the art Models such as T5 Model, BERT and GPT3 for identifying radicalized tweets.

4. Detecting the sentiments of users through tweets using Python Libraries such as NLTK/CoreNLP.

Project Implementation Method

Our aim is to propose and develop an effective radicalism detection solution. After acquiring labelled dataset from Kaggle , the data has been cleaned through different pre-processing techniques: removing hyperlink, tags and RTs, null and short tweets removal, removing punctuation, stop-words removal, word lemmatization and Feature Extraction. To classify tweets displaying extremist tendencies, different Machine Learning Algorithms such as Support Vector Machine, Deep NN, Recurrent NN ,BERT T5 and Random Forest will be used. The algorithm with the best accuracy is implemented. For the identification of emotions, opinions, sentiments showing allegiance to extremist causes in tweets, Python Libraries based on natural language processing such as NLTK/CoreNLP have been used for sentiment analysis to be categorized into positive, negative or neutral sentiments

Benefits of the Project

• Identification of tweets containing radical content related to religious cyber-extremism.

• To measure the performance of Machine Learning models such as Naïve Bayes, Support Vector Machine, Random Forest, and Deep Neural Networks in identifying radicalized tweets.

• Performing sentiment analysis on tweets; identifying users with radicalized mindset and mining their emotions through their tweets.

• To mine emotions through tweets using Python’s Libraries for natural language processing

Technical Details of Final Deliverable

Our product will be a piece of software that can be implemented in other social platforms. The Machine Learning algorithms such XGBoost Random Forest and Support Vector Machine can be trained using CPU but Deep Learning Models such as Vanilla RNN ,Transformers and Transfer learning of BERT and T5 may take months to get trained on CPU , so for that purpose RTX Graphical Processing Unit is required for the fast training of models as RTX GPU contains Tensor cores which aligns with Tensor Flow Deep Learning Library of Google and  make training fast on GPUs.

Final Deliverable of the Project

Software System

Core Industry

IT

Other Industries

Media

Core Technology

Artificial Intelligence(AI)

Other Technologies

Big Data

Sustainable Development Goals

Industry, Innovation and Infrastructure, Reduced Inequality, Peace and Justice Strong Institutions

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com