Contextualizing Internet Memes Using Knowledge Graphs: An Analytical Study

This research introduces an analytical methodology for contextualizing internet memes through knowledge graphs to improve their semantic clarity. A dataset of 1,500 memes was amassed from Instagram, Twitter/X, and YouTube, thereafter undergoing preprocessing, OCR-based text extraction, and the production of multimodal embeddings via CLIP. Entities, relationships, and emotions were extracted to create a Meme Knowledge Graph (MKG) in Neo4j, facilitating structured semantic representation and contextual analysis. The MKG attained a mean Semantic Richness Score (SRS) of 0.64, indicating substantial enhancement in meme comprehension, especially with political and situational humor. Graph-based contextualization enhanced downstream analytical tasks, resulting in increases of 15%, 19%, and 21% in classification accuracy, sentiment interpretation, and misinformation detection, respectively, compared to baseline models. The findings validate that knowledge graphs offer a strong and comprehensible method for modelling the socio-cultural and affective aspects of online memes.