Latent-Optimized Adversarial Neural Transfer for Sarcasm Detection

TL;DR

This paper introduces a latent-optimized adversarial transfer learning method that enhances sarcasm detection across datasets, achieving significant performance improvements over existing approaches.

Contribution

It proposes a generalized latent optimization strategy to better balance multiple loss functions in adversarial transfer learning for sarcasm detection.

Findings

Achieves 10.02% absolute performance gain on iSarcasm dataset.

Outperforms transfer learning and meta-learning baselines.

Improves training stability and transfer effectiveness.

Abstract

The existence of multiple datasets for sarcasm detection prompts us to apply transfer learning to exploit their commonality. The adversarial neural transfer (ANT) framework utilizes multiple loss terms that encourage the source-domain and the target-domain feature distributions to be similar while optimizing for domain-specific performance. However, these objectives may be in conflict, which can lead to optimization difficulties and sometimes diminished transfer. We propose a generalized latent optimization strategy that allows different losses to accommodate each other and improves training dynamics. The proposed method outperforms transfer learning and meta-learning baselines. In particular, we achieve 10.02% absolute performance gain over the previous state of the art on the iSarcasm dataset.