Abstract:
"Cyberbullying has become an important concern in today's culture due to the increasing usage
of social media. With recent events such as Covid-19 and the economic crisis taking place, the
number of social media users in Sri Lanka has grown to the top on a vast scale. Despite
investing millions of dollars to solve this issue, prominent social media sites such as Facebook,
Instagram, and Twitter remained struggling to make a significant shift on their platforms.
Social media platforms have grown in popularity as a tool of communication for languages
with low resources. However, the detection of cyberbullying in various transliterated forms of
such languages is limited. Due to a lack of expertise of such languages, it is difficult for
platforms to respond quickly during riots on social media platforms. This study utilizes an
automated technique to detect cyberbullying in Romanized Sinhala, a low-resource language
which will help the community to overcome this major concern.
With this in place, cyberbullying in the Singlish (Sinhala & English) mixed-code language has
been highlighted as a big concern in the Sri Lankan community. In respond to this problem, a
cyberbullying detection system for Singlish was proposed, which utilized a hybrid algorithm
combining Naive Bayes and MLP Classifier. The suggested technology aims to identify
cyberbullying automatically on social media platforms, hence lessening the detrimental impact
of cyberbullying on Sri Lankan society. The goal of this study is to develop a critical tool for
identifying cyberbullying on social media using the proposed hybrid algorithm for Singlish
language.
This study provides a thorough examination and critique of past studies and technologies in
order to support the suggested solution. Furthermore, the author goes into detail about the
testing and evaluation procedures that were used. Based on the results of the assessments, the
author considers that the proposed system is a very effective solution to the current problem"
