Understanding Sarcasm in Social Media

This project explores how sarcasm is expressed and understood across text, images, and videos on social media. It develops computational approaches to detect sarcasm and examines how context, multimodality, and user interaction shape its meaning.

Outcome: ICCBD 18, ICMI 18, HCC 19, TransAI 19, and TransAI 19

Overview

Sarcasm is a common but complex part of human communication. People often say one thing but mean the opposite, making it difficult for both humans and AI systems to interpret correctly.

This project explores how sarcasm is expressed on social media and how it can be detected using computational methods. It looks beyond text to include images, videos, and user interactions, showing that sarcasm is not just about words—it depends on context and multiple signals.

Approach

This research studies sarcasm across different types of data and platforms:

  • Text and multimodal analysis: Combining text, images, and metadata from social media posts
  • Computer vision methods: Using CNN-based models to detect sarcasm from visual cues in images
  • User interaction modeling: Analyzing reactions, comments, and engagement patterns as signals of sarcasm
  • Video analysis: Studying attention patterns in sarcastic videos using gaze data and AI models

Key Findings

  • Sarcasm cannot be reliably detected from text alone—it often depends on visual and contextual cues
  • User interactions (e.g., reactions, comments) provide important signals for understanding sarcastic intent
  • Different media types (images, videos) reveal distinct patterns of how sarcasm is communicated
  • In longer content like articles, tone and storytelling style help distinguish satire from misleading information

Contributions

  • Early demonstration of multimodal sarcasm detection across text, images, and videos
  • Introduces the role of user interaction and context in understanding online communication
  • Develops datasets and models for sarcasm detection in social media environments
  • Connects sarcasm to broader questions of misinformation, interpretation, and meaning in digital content