NLP Framework to Safeguard Youngsters Online Using Advanced Transformer-Based Models

Abstract

Ensuring the safety and well-being of youngsters in online environments has become increasingly challenging, particularly with the rise of harmful and inappropriate content on social media platforms. This research study focuses on developing a natural language processing (NLP) framework designed to monitor online interactions and identify online harmful conversations. The framework integrates emotion detection with an advanced age verification model to analyze communication patterns and detect inappropriate behavior. When repeated sexually inappropriate behavior exceeds a predefined threshold, the system responds by blocking accounts and issuing notifications to guardians. The NLP framework was evaluated using two traditional machine learning algorithms alongside two advanced models, namely, BERT and RoBERTa, to assess their effectiveness in detecting harmful interactions. These models were trained and tested on datasets containing emotional patterns and real-world social media conversations. The proposed model’s best accuracy result was 95.15%, which shows great promise in addressing inappropriate behavior during a conversation. However, several challenges were identified, including managing imbalanced data and the substantial computational resources required for model training. Despite these limitations, the framework can be used to enhance online safety for youngsters.

Received: 21 March 2025 | Revised: 15 October 2025 | Accepted: 3 December 2025

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data that support the findings of this study are openly available in Kaggle at https://www.kaggle.com/datasets/mrmorj/hate-speech-and-offensive-language-dataset, https://www.kaggle.com/datasets/debarshichanda/goemotions, and https://www.kaggle.com/datasets/bhavikjikadara/emotions-dataset.

Author Contribution Statement

Hisham AbouGrad: Conceptualization, Methodology, Validation, Investigation, Data curation, Writing — original draft, Writing — review & editing, Visualization, Supervision, Project administration. Sankar Santhosh: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing — original draft, Visualization. Salem Alsaid: Writing — review & editing, Visualization.