Sarcasm Detection¶
Sarcasm is a form of non-literal language where the intended meaning differs from the literal meaning, fulfilling a social function in discourse. Sarcasm detection is challenging in NLP because it requires understanding linguistic context, user behavior, and often cultural or implicit knowledge.
Unlike irony (which can be detected through linguistic markers), sarcasm is fundamentally tied to social relationships and intent. A statement intended sarcastically by one person may be interpreted literally by another, making the distinction between perceived sarcasm (how an audience interprets an utterance) and intended sarcasm (what the author meant) both important and difficult.
Key papers¶
- Perceived and Intended Sarcasm Detection with Graph Attention Networks — graph attention networks modeling user context and social relationships on Twitter
- A Multi-Modal Method for Satire Detection using Textual and Visual Cues — multi-modal satire detection (related, distinct phenomenon of deliberate deception)
Related topics¶
- Sentiment Analysis (affected by sarcasm; detecting sarcasm improves sentiment prediction)
- Stance Detection (sarcasm can invert apparent stance; related annotation and evaluation challenges)
- Irony Detection (linguistic cousin; irony markers are text-internal, sarcasm markers are social)
- Social Media Analysis (sarcasm prevalent on Twitter, Reddit, informal platforms)
- User Context (sarcasm interpretation depends on author history and audience relationship)