Deep Multi-Task Model for Sarcasm Detection and Sentiment Analysis in Arabic Language

TL;DR

This paper presents a deep multi-task learning model using BERT for simultaneous sarcasm detection and sentiment analysis in Arabic, effectively leveraging shared knowledge to improve performance.

Contribution

It introduces an end-to-end multi-task model that enables knowledge sharing between sarcasm detection and sentiment analysis tasks in Arabic.

Findings

The multi-task model outperforms single-task models on both tasks.

Shared representations improve sarcasm detection accuracy.

Shared representations enhance sentiment analysis performance.

Abstract

The prominence of figurative language devices, such as sarcasm and irony, poses serious challenges for Arabic Sentiment Analysis (SA). While previous research works tackle SA and sarcasm detection separately, this paper introduces an end-to-end deep Multi-Task Learning (MTL) model, allowing knowledge interaction between the two tasks. Our MTL model's architecture consists of a Bidirectional Encoder Representation from Transformers (BERT) model, a multi-task attention interaction module, and two task classifiers. The overall obtained results show that our proposed model outperforms its single-task counterparts on both SA and sarcasm detection sub-tasks.