An Efficient Deep Learning Approach to Deal with Cyberbullying

Now a days Internet become more Integral part of World particularly social media platforms, such as Instagram, Facebook & What’s app etc. which led to emergence of cybercrime. Cyberbullying is a special way of bullying which emerged with development of social networks. People are bullyingother people by commenting or by posting toxic commentor embarrassing picture which lead to cause harm to Victim reputation. It is very common among teenagers. Stating to the above challenge we have pitched a deep learning based modelto identify bullied text in meme data, our model performed well in categorizing the text in to toxic comment 10 %, obscene in between 5-10 %, insult is about 5 %, severe toxic & identity hate is almost between 2 % & threat is 0-5 %.

I. Introduction

Cyberbullying is the deliberate, severe harm that is committed through a computer or other electronic equipment [1], The main reason behind that cyberbullying increasing now a days because of mass of people are involved in social media platform, where a user can engage with large numbers of people. Culprit can easily disturb the users without his/her physical presence & tried to black mail people by sharing embarrassing picture which may deteriorate the reputation of victim. Here in our project, we are using text and image processing for detection of Cyberbullying. It automatically detects the cyberbullying by leveraging visual data. Here we are reporting visual features compliments & textual traits that are really beneficial in improving predictive result. We are going to do systematic analysis on Cyberbullying by referring different papers [2], Cyberbullying is a phenomenon that is described as someone purposefully, deliberately, and repeatedly harming another person by offensive remarks, messages, or other types of social aggression using different digital technologies, though In the field of social science research, there has been a lot of interest in studying cyberbullying, but there has been far less attention paid to it in the field of automatic detection research, In the realm of automatic detection research, it has gotten far less attention. Current research in this field regards cyberbullying detection as a two or multi- class classification problem under supervised learning [3] [29], adequate examples for both positive and negative training are required in such instances for efficient categorization performance. Social networking has grown in popularity in recent years. Additionally, social networking websites are quite useful for meeting new people [4], However, popularity of social networking has increased. People are using these communities in illegal and unethical ways. People, particularly teenagers and young adults, are finding new ways to attack one another on the and teenagers are coming up with new methods to harass one another online. Bullying is not a new phenomenon, and when digital technology became a primary form of communication, cyberbullying evolved. Social media platforms have the benefit of allowing you to communicate with anyone, at any time. Blogs, social networking sites like Facebook, and instant messaging programmes like WhatsApp are a few examples allow you to engage with anybody and at any time [5], The reel world functions as a web-connected network of people from all around the world. The medium that was supposed to be used to connect and communicate with people has instead become a place where people, particularly teenagers and young adults, bully one another. Though social media allows people to interact, it also exposes them to potentially dangerous situations such as aggressive cyberbullying, which can lead to sadness and suicide thoughts [6].