Finetuning for Sarcasm Detection with a Pruned Dataset

TL;DR

This paper improves sarcasm detection by fine-tuning a RoBERTa model with a significantly reduced Reddit dataset, achieving near state-of-the-art performance on the iSarcasm dataset.

Contribution

It introduces a pruned, smaller version of the Reddit dataset to enhance sarcasm detection models without extensive data requirements.

Findings

Achieved within 0.02 F1 of state-of-the-art performance.

Used a dataset 100 times smaller than previous training data.

Demonstrated effective fine-tuning with reduced dataset size.

Abstract

Sarcasm is a form of irony that involves saying or writing something that is opposite or opposite to what one really means, often in a humorous or mocking way. It is often used to mock or mock someone or something, or to be humorous or amusing. Sarcasm is usually conveyed through tone of voice, facial expressions, or other forms of nonverbal communication, but it can also be indicated by the use of certain words or phrases that are typically associated with irony or humor. Sarcasm detection is difficult because it relies on context and non-verbal cues. It can also be culturally specific, subjective and ambiguous. In this work, we fine-tune the RoBERTa based sarcasm detection model presented in Abaskohi et al. [2022] to get to within 0.02 F1 of the state-of-the-art (Hercog et al. [2022]) on the iSarcasm dataset (Oprea and Magdy [2019]). This performance is achieved by augmenting iSarcasm with a pruned version of the Self Annotated Reddit Corpus (SARC) (Khodak et al. [2017]). Our pruned version is 100 times smaller than the subset of SARC used to train the state-of-the-art model.