A Large Self-Annotated Corpus for Sarcasm

2017/04/19 by Mikhail Khodak, Nikunj Saunshi, Khodak, Mikhail +3 · 1 voice · 7 citations
#cs.CL #cs.AI #cs.LG

paper · pdf · doi:10.48550/arxiv.1704.05579

Abstract

We introduce the Self-Annotated Reddit Corpus (SARC), a large corpus for sarcasm research and for training and evaluating systems for sarcasm detection. The corpus has 1.3 million sarcastic statements -- 10 times more than any previous dataset -- and many times more instances of non-sarcastic statements, allowing for learning in both balanced and unbalanced label regimes. Each statement is furthermore self-annotated -- sarcasm is labeled by the author, not an independent annotator -- and provided with user, topic, and conversation context. We evaluate the corpus for accuracy, construct benchmarks for sarcasm detection, and evaluate baseline methods.

Cited by

• ContextLeak: Auditing Leakage in Private In-Context Learning Methods
• SocialNLI: A Dialogue-Centric Social Inference Dataset
• Transfer Learning via Lexical Relatedness: A Sarcasm and Hate Speech Case Study
• SeLeRoSa: Sentence-Level Romanian Satire Detection Dataset
• Large Language Models for Subjective Language Understanding: A Survey
• Irony punctuation [wikipedia]
• Tone indicator [wikipedia]

Discussions

• A Large Self-Annotated Corpus for Sarcasm [hn, 105 points, 29 comments]

Related

• An Evaluation of State-of-the-Art Large Language Models for Sarcasm Detection
• Sarcasm Detection on Reddit Using Classical Machine Learning and Feature Engineering
• Automatic Sarcasm Detection: A Survey
• Leveraging Large Language Models for Sarcastic Speech Annotation in Sarcasm Detection
• Computational Sarcasm Analysis on Social Media: A Systematic Review