이둠밈, 신영주,

Vol. 35, No. 2, pp. 313-321, 4월. 2025
10.13089/JKIISC.2025.35.2.313, Full Text:
Keywords: Profanity Detection, Malware visualization, Machine Learning, Content Moderation
Abstract

This paper explores a novel approach to detect profanity in Korean textual data by converting text to HSV scale images. Traditional natural language processing (NLP) methods often struggle with contextual nuances and the use of non-standard language in online communication. By adopting image visualization-based analysis used in traditional malware detection, this research aims to discover structural and visual patterns associated with profanity that are overlooked by standard text-based classifiers. By utilizing different image classification artificial intelligence architectures on a dataset of Korean sentences converted to HSV scale images, we aim to demonstrate the potential of image-based analysis to complement traditional text analysis in automated content moderation systems. Experimental results show that the proposed model achieves 97.42% accuracy.

Statistics
Cite this article