This research introduces a keyword-oriented multimodal framework to identify euphemisms in online content, critical for security monitoring and content moderation.

• Addresses the challenge of detecting hidden meanings in text with supporting images/audio
• Focuses on identifying euphemisms in illicit contexts (drugs, weapons, sexuality)
• Pionеers multimodal analysis for euphеmism dеtеction, a gap in current research
• Enables more effective content moderation systems for digital platforms

The approach directly enhances security applications by helping identify coded language in underground markets, improving automated content moderation, and supporting law enforcement in monitoring illicit activities.

Keyword-Oriented Multimodal Modeling for Euphemism Identification