The evolution of the internet gave us capabilities to be interconnected by each other over the globe, but every entity that has evolved in mankind also comes with some demerits. The evolution of the web from just being a simple search index to blogging and social networking, it gives everyone a freedom of speech which makes a chaotic situation about polling and changing perceptions of the people it may cause a social dilemma where an individual has unconsciously maintained 2 personalities. However, the evolution in computer analytical field gave us the privilege of using various available tools and techniques for analyzing the social media and create a policy on the freedom of speech, Our paper focuses on the solution for regulating the policy and guidelines of the social media platform with the help of machine learning techniques and natural language processing. Nowadays there are more and more people who are having a perception of harassing others on social networks, these people may have a different behavior towards others in real life. We have developed a solution to filter out the content on social media, first filtering out the slangs and harassing comments, then it can be used on the blog posts, or even on the images using computer vision techniques. It is necessary to use the language pattern and grammar to maintain a high order overview by which toxicity can be judged. The solution in this particular situation are approached best by Neural Networks, we are using a better Neural Network which is more effective and replace the sequential nature of the neural network, i.e. BERT and it makes use of transformers which is just an attention-based mechanism that can learn contextual relations between words in a text, to identify the text we will be using a Convolutional Neural Network which can help in NLP.

Neural Network Convolution Neural Network NLP BERT