Scholarlink

<p>Educational institutions serve not only as centers for academic research but also as spaces that reflect diverse cultural and social identities in contemporary learning contexts. Regretfully, the presence of hate speech and hate symbols poses a serious threat to both the physical and digital environments within these institutions. Such behaviors can disrupt students' personal and academic lives, contribute to psychological distress—such as anxiety, depression, and diminished self-esteem—and impede efforts to cultivate inclusive learning communities by fostering an atmosphere of fear. Despite the urgency of the issue, current methods for regulating hate speech and symbols often lack scalability and real-time responsiveness. Therefore, there is a critical need for intelligent systems that can efficiently identify and respond to harmful content while also safeguarding individual rights and promoting a positive educational environment. This study addresses this gap by proposing an Artificial Intelligence-inspired IoT-fog computing framework designed to detect hate speech and symbols, specifically those found in posters within educational institutions. The proposed system employs the YOLOv8 model to identify prohibited symbols and weapons and utilizes EasyOCR for text extraction. Subsequently, a modified DistilBERT model performs the contextual evaluation of the detected objects and text. The experimental results, conducted on sample images from the dataset, indicate strong performance, with an F1 score of 0.87, a recall rate of 0.90, and an accuracy of 0.94, along with other favorable evaluation metrics. These findings demonstrate that the proposed system outperforms existing state-of-the-art methods in detecting and alerting users to potentially harmful content. Implementing such an AI-driven framework offers a practical and effective solution for institutional administrations to address challenges related to hate propagation, violence incitement, and mob instigation. Consequently, the framework contributes to the creation of safe, intelligent campuses, reduces the risk of adverse incidents, and supports academic success within an environment of mutual respect.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial intelligence assisted framework for detecting offensive posters on the premises of educational institutions

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Munish Saini,
  • Eshan Sengupta,
  • Harsh Sharma,
  • Suraaj Thakur

摘要

Educational institutions serve not only as centers for academic research but also as spaces that reflect diverse cultural and social identities in contemporary learning contexts. Regretfully, the presence of hate speech and hate symbols poses a serious threat to both the physical and digital environments within these institutions. Such behaviors can disrupt students' personal and academic lives, contribute to psychological distress—such as anxiety, depression, and diminished self-esteem—and impede efforts to cultivate inclusive learning communities by fostering an atmosphere of fear. Despite the urgency of the issue, current methods for regulating hate speech and symbols often lack scalability and real-time responsiveness. Therefore, there is a critical need for intelligent systems that can efficiently identify and respond to harmful content while also safeguarding individual rights and promoting a positive educational environment. This study addresses this gap by proposing an Artificial Intelligence-inspired IoT-fog computing framework designed to detect hate speech and symbols, specifically those found in posters within educational institutions. The proposed system employs the YOLOv8 model to identify prohibited symbols and weapons and utilizes EasyOCR for text extraction. Subsequently, a modified DistilBERT model performs the contextual evaluation of the detected objects and text. The experimental results, conducted on sample images from the dataset, indicate strong performance, with an F1 score of 0.87, a recall rate of 0.90, and an accuracy of 0.94, along with other favorable evaluation metrics. These findings demonstrate that the proposed system outperforms existing state-of-the-art methods in detecting and alerting users to potentially harmful content. Implementing such an AI-driven framework offers a practical and effective solution for institutional administrations to address challenges related to hate propagation, violence incitement, and mob instigation. Consequently, the framework contributes to the creation of safe, intelligent campuses, reduces the risk of adverse incidents, and supports academic success within an environment of mutual respect.