Comparison of Multilingual and Bilingual Models for Satirical News Detection of Arabic and English

TL;DR

This paper compares multilingual and bilingual models for satirical news detection in Arabic and English, demonstrating that structured reasoning via chain-of-thought prompting improves model performance, especially with Jais-chat.

Contribution

It introduces a comparative analysis of multilingual and bilingual models for satire detection and highlights the effectiveness of chain-of-thought prompting in enhancing model accuracy.

Findings

Jais-chat with CoT prompting achieved 80% F1-score in English.

CoT prompting significantly improves satire detection performance.

Structured reasoning is crucial for complex NLP tasks like satire detection.

Abstract

Satirical news is real news combined with a humorous comment or exaggerated content, and it often mimics the format and style of real news. However, satirical news is often misunderstood as misinformation, especially by individuals from different cultural and social backgrounds. This research addresses the challenge of distinguishing satire from truthful news by leveraging multilingual satire detection methods in English and Arabic. We explore both zero-shot and chain-of-thought (CoT) prompting using two language models, Jais-chat(13B) and LLaMA-2-chat(7B). Our results show that CoT prompting offers a significant advantage for the Jais-chat model over the LLaMA-2-chat model. Specifically, Jais-chat achieved the best performance, with an F1-score of 80\% in English when using CoT prompting. These results highlight the importance of structured reasoning in CoT, which enhances contextual understanding and is vital for complex tasks like satire detection.