Table of Contents
Abstract
Social Network has become one of the important aspects in our day-to-day life for students, employees, etc.; for communication experience, stress relief and so on. As social media increasing massively a serious problem i.e, cyber-bullying has merged mostly affecting youngsters and teenagers at worst causing their suicidal attempts. This may lead to a negative impact on people’s life. Cyber-bullying is an action performed knowingly or willingly by every individual or group of individuals through messages, pictures from their devices that can cause pain (hurt, embarrass) to the victim. By Machine learning techniques we can automatically detect the cyber-bullying content and the pattern of the language used in social media which can create a safer environment for the people using social media.
A new machine learning technique is proposed here for cyber-bullying detection. This technique(method) named Semantic-Enhanced Marginalized Auto-Encoder(smSDA) developed which is an extension of the deep learning model. This technique has experimented on two sites i.e.; Twitter and Myspace. This performed for detection as well as blocking the accounts of the bullying persons on the social network. The result of this method is greater compared to the following text representation machine learning methods.
Introduction
Nowadays, as there is an increase in population the use of social media is increasing day by day. Social media has become a basic activity in people’s lives, especially for children and teenagers. The users of social networking sites are growing rapidly over millions. Most of the people are spending 17 hours per week, and we see in some cases spending 40 hours and more per week online by communicating with friends, interacting with others irrespective of their location. This caused a side effect that evolved a serious problem called cyber-bullying.
127 our experts are availableright now
Cyber-bullying is a problem raised by one single individual or by a group who performs activities intentionally or non-intentionally that may lead to hurt the victim. Bullying may be in several ways in sites of social network i.e.; can be of messages, comments, pictures, posts, etc. This causes the victim to feel depressed and affect his/her personal life which may lead to suicidal attempts. Cyber-bullying is different from traditional bullying which occurs at any place. Some of the places where cyber-bullying occurs are:
- Networking sites such as Facebook, Whats App, Twitter, Instagram, Snap-chat and so on
- Text- based messages from devices i.e.; SMS
- Emails
Cyber-bullying tactics:
The activities performed on the sites of the social network are: - Sending messages containing bullies
- Posting comments on others timeline posts
- Uploading pictures or videos without their permission
The word embeddings are suggested for taking out the bullying words used. Through smSDA instructing we attempt to renovate the attributes of bullying from other words. Details of correlation are located by smSDA renovate attributes of bullying from normally used words, used to detect the bullying messages not containing bullying words.
Experimental Results
The main aim of this project is to detect the cyber-bullying words and avoid them. After detection, it blocks the user’s account and denies s access. In this firstly the user has to register by entering his/her details. After registering the user can log in his/her account respectively. The user can perform many actions using their account. Users can share posts through their account.
They can send messages where public and private options are available to have a chat. Users can send/receive friend requests to other users. They can also search for the other profiles of the users and accept the requests they get from other users. If in case any user using bullies in their messages or posts immediately he gets blocked by the admin and his account id will be added in the blocked list. The details of the user who is using bullying words will be detected and sent to the admin. Admin can add categories of words i.e.; illegal/bad. Lastly, the user’s detail who used bullies won’t appear to anyone as their denial of access by admin.
Conclusion
As, the paper mainly discusses the text-based cyber-bullying detection problem, where robust and discriminative represents the structure of informational messages that are critical for an effective detection system. As the main goal of this project is the detection of bullies on the social network. This approach is used for detection as well as avoidance of cyber-bullying words. During implementation, it detects and blocks the bullying user. It also denies the access of the user using bullying words. This makes users feel comfortable using the social network.
