Home
Research Highlight
MemeMatch, Dual-Context Multimodal Meme Dataset and Retrieval
Dual-context pipeline: local context (OCR overlay text + title) and global context (template semantics).
MemeMatch is a large-scale multimodal meme dataset and retrieval system for studying how meaning, intent, and emotion connect to online engagement and virality. It includes rich annotations such as emotion vectors, topics, and usage-intent labels, built through a dual-context pipeline:
- Local context: OCR overlay text plus post title
- Global context: template semantics and visual meaning
On top of this representation, MemeMatch supports intent-aware search (for example, “sarcastic memes about college”) and image-based retrieval, enabling analyses of how memes are reframed across communities and templates, and how those shifts relate to engagement.
➡️ See: Publications · CV
News
-
Mar 2026: MemeMatch got accepted at ICWSM 2026!
-
May 2025 - Aug 2025: Internships at Citywide Classroom and Re-Volt InnovationsJan 2025 - May 2025: Research abroad at AIT Budapest and HSDSLab
-
Jun 2024 - Aug 2024: Summer Undergraduate Math and Statistics Accelerator (SUMSA), IMSI, The University of Chicago.
-
May 2023 - Aug 2023: Machine Learning Research Assistant, Department of Mathematics and Computer Science, Wabash College.