An Exploration of the Potential Use of NLP Techniques in Social Media Platforms for Profiling Potential Paedophiles

EasyChair Preprint 9265

Abstract

Social Medial Platforms are now widely used by children and teenagers, where privacy, safety and their security online are often compromised. This has led to children being exposed to paedophiles while they are innocently using different types of social media platforms. The paedophiles lure children by creating simple conversation starters in the form of relatable simple chats, which can be the first step to online child abuse. Although there has been an abundance of tools trying to protect children from online child abuse, the challenge remains far from being solved yet. This paper focuses on how NLP techniques can be used on creating an effective way of protecting children against child online abuse while they are using online social media platforms. Informed from an extensive study of relevant literature and implementation cases, this research presents some of the key insights identified. Among the key findings are: Recent surveys confirming a significant increase in online child abuse even though there are some profiling tools available. The main conclusion drawn from this paper were that current approaches to fight online child abuse are deficient because they fail to accurately identify the words that paedophiles use when communicating with children on social media platforms. This paper argues for the use of an NLP tool, that can detect potential paedophiles using the common keywords they use when communicating with children online, hence the proposal towards curbing child online abuses smartly.

Keyphrases: Artificial Intelligence, Natural Language Processing, Paedophiles, social network sites

@booklet{EasyChair:9265,
  author    = {Timothy Takawira Dongo and Attlee M Gamundani and Mercy Chitauro},
  title     = {An Exploration of the Potential Use of NLP Techniques in Social Media Platforms for Profiling Potential Paedophiles},
  howpublished = {EasyChair Preprint 9265},
  year      = {EasyChair, 2022}}