Transformer-based Context-aware Sarcasm Detection in Conversation Threads from Social Media

TL;DR

This paper introduces a transformer-based model that leverages entire conversation threads to improve sarcasm detection accuracy on social media data, achieving state-of-the-art results on Twitter and Reddit datasets.

Contribution

It presents a novel context-aware sarcasm detection approach using deep transformer layers to incorporate conversation context, outperforming baseline models.

Findings

Achieved 79.0% F1-score on Twitter dataset.

Achieved 75.0% F1-score on Reddit dataset.

Outperformed 35 other systems in shared task.

Abstract

We present a transformer-based sarcasm detection model that accounts for the context from the entire conversation thread for more robust predictions. Our model uses deep transformer layers to perform multi-head attentions among the target utterance and the relevant context in the thread. The context-aware models are evaluated on two datasets from social media, Twitter and Reddit, and show 3.1% and 7.0% improvements over their baselines. Our best models give the F1-scores of 79.0% and 75.0% for the Twitter and Reddit datasets respectively, becoming one of the highest performing systems among 36 participants in this shared task.