SafeSpace: Cyberbullying Detection using Deep Learning
ABSTRACT
Cyberbullying is the use of technology as a medium to bully someone. Although it has been
an issue for man-years, the recognition of its impact on young people has recently increased.
Social networking sites provide a fertile medium for bullies, and teens and young adults who
use these sites are vulnerable to attacks. Through machine learning, we can detect language
patterns used by bullies and their victims, and develop rules to automatically detect cyberbul-
lying content. Over the last decade, social media has acquired a lot of traction, both positively
and negatively way. With the fast growth of social networking, People can communicate with
one other via platforms and websites. Directly with no cultural or economic barriers While
There have been several advantages to using social media, yet there are [Link] negative so-
cietal effects One such issue that has arisen is Hate speech has been more prevalent in recent
years. Hateful speeches essentially the use of rude and abusive words. using social media It
might relate to anybody or something specific. a collection of people who have common in-
terests. In this study, we introduced our approach to dealing with hate speech and, to a con-
siderable part, decreasing it. People express their hatred and rage on social media right in -
stantly, which hurts the sentiments of others. To eradicate hate speech, we dug deep into nat-
ural language processing and employed several machine learning models to choose which
one to deploy based on accuracy.
Keywords: Natural Language Processing, Classification Technique and Logistic Regression,
CNN and Naïve Bayes Classification
INTRODUCTION
Cyberbullying is defined as the use of offensive information such as harassment, insult, and
hatred in messages sent or posted via OSN services with the intent of intentionally injuring
people emotionally, mentally, or physically Social media has become the primary means of
communicating ideas throughout the world in recent years. According to reference. Social
media is a trustworthy medium, but because there is so much material shared and debated, it
might be impossible to review a remark in a single sitting, resulting in a rise in hate speech.
People may now connect directly with each other across cultural and economic boundaries
because to the fast growth of networking through social platforms and websites. Hate speech
may also be defined as an emotional idea. We may define hate speech as the use of rude and
abusive statements on social media. It can relate to any individual or a specific group of indi-
viduals who have similar interests. In this study, we introduced our approach to dealing with
hate speech and, to a considerable part, decreasing it.
Hate speech has escalated dramatically in recent years. In reality, it has worsened since the
COVID-19 pandemic-related shutdown because all work and communication has been done
online since then. People of different ages, cultures, and hobbies are increasingly using social
media platforms such as Twitter, Facebook, and Instagram. These are some of the sites where
hate speech has been reported.
These platforms provide an open stage for people to voice their opinions and share or convey
their thoughts and messages across the world but the enormous number of posts and ex-
change of messages makes it almost impossible to keep their content in control. Facebook has
a set of community guidelines in place to deal with abuse, online bullying and illegal behav -
ior, sexual assault and violence on public figures. Similar to Facebook, Twitter too has some
guidelines that would aid someone who is a victim to social abuse.
Hate speech not only causes turmoil and friction among diverse groups, but it also causes
real-world problems. This study defines hate speech and provides several instances of how
hate speech happens. It is mostly concerned with dealing with hate speech on Twitter. The
data pre-processing is shown next. It goes on to detail the methodologies employed on the
dataset, such as sentiment analysis, semantic analysis, the Unigram feature, and pattern ex-
traction. It includes graphs and tables that demonstrate the precision and accuracy attained for
various models.
EXISTING METHOD
In existing system, we have used ANN to classify the sentiment which is not that much effi-
cient. This technique consumes tons of time for analysing the content and also the accuracy
obtained by the model is less compared to other algorithms.
Disadvantages:
• Low accuracy.
• Time consuming.
• High complexities.
PROPOSED METHOD
Proposed several machine learning models to classify whether the speech that we are going to
classify is a hate one or not, but none have adequately addressed this misdiagnosis problem.
Also, similar studies that have proposed models for evaluation of such performance classifi-
cation mostly do not consider the heterogeneity and the size of the data Therefore, we pro-
pose a Natural Language Processing and Logistic Regression techniques and Naïve Bayes
classifier to predict whether it is a hate speech or not.
Fig 1. Block diagram of proposed method
Advantages:
High accuracy.
Time Saving.
Low complexities.
High reliability.
HARDWARE AND SOFTWARE REQUIREMENTS
H/W Configuration:
Operating system : Windows 7 or 7+
RAM : 8 GB
Hard disc or SSD : More than 500 GB
Processor : Intel 3rd generation or high or Ryzen with 8 GB Ram
S/W Configuration:
Software’s : Python 3.6 or high version
IDE : PyCharm.
Framework : Flask