An Intelligent Deep Learning Framework For Social Media Bot Detection Using Transformer Based Model

Abstract

Nowadays, Social Media Platforms Have Emerged As One Of The Primary Modes Of Communication Where Memes And Comments Are Used Extensively For Spreading Ideas, Humor, And Thoughts. This Increasing Trend, However, Has Been Accompanied By The Speedy Proliferation Of Offensive Language, Hateful Rhetoric, And Cyberbullying, All Of Which Harm Individuals And Their Communities On Social Media Platforms. The Current Solutions For Moderating Such Content Rely Mainly On Manual Reporting And Simplistic Techniques Like Keyword Filters, Which Are Inefficient Because They Do Not Comprehend Context And Are Ineffective Because They Lack Speed. Memes, Especially, Have Text Along With Graphics And Images, Which Complicates Efforts To Automatically Detect Any Kind Of Harmful Content Contained In Them. As A Solution To This Problem, The Proposed Project Will Design A Smart System For Classifying Memes. In The Proposed System, Optical Character Recognition (OCR) Technology Is Applied To Recognize Text From Memes, Which Is Then Analyzed With NLP Technologies Like Tokenization, Stemming, And Stop Word Removal. The Sentiment Of The Posts Is Classified Using The VADER Algorithm Into Three Categories: Positive, Negative, And Neutral. At The Same Time, Deep Learning Image Classifier Is Used To Determine If There Are Any Unsuitable Pictures In Posts. Based On The Output, The System Can Automatically Remove Harmful Content, Issue Notifications And Warnings, As Well As Keep Track Of User Activity, Thus Blocking Users Who Frequently Violate The Community Standards.

Keywords