An Opinion Mining Based Approach for Filtering Potentially Extremist and Radicalized Content

An Opinion Mining Based Approach for Filtering Potentially Extremist and Radicalized Content

Jaya Bhatnagar1, Vijay Sankar Mishra2, Dinesh Kumar Gupta3

Abstract: With the emergence of social media as a common platform for communication among different people and communities at large, the chances for malicious usage of social media platforms for malicious activities has also increased manifold. One such malicious activity is spreading radical content over social media platforms due to the ease of sharing among several individuals and groups. The challenging aspect though for social media agencies or security agencies is the screening of humongous amounts of data to detect radical content. With no clear boundary to demarcate radical and non-radical content, the classification problem becomes challenging as the data size increases.[1] The proposed work presents an artificial intelligence based technique for detection of radical content. The proposed approach uses the concept of dictionary learning to train a Bayesian Regularized artificial neural network. The performance evaluation parameters are the number of iterations, absolute time and the accuracy. It has been shown that while the proposed system attains a classification accuracy of 97% compared to 89% of previous work.

Keywords:-  Radical Content, Counter-terrorism, social networks, text analysis, Bayesian Regularization, accuracy.