Subjectivity and Sentiment Analysis of Modern Standard Arabic and Arabic Microblogs

TLDR

SSA research has largely ignored Arabic, despite extensive work in other languages. The study targets SSA of Modern Standard Arabic news and Arabic microblogs, highlighting challenges in microblog sentiment analysis. The authors extended an Arabic SSA lexicon via a random graph walk on Arabic‑English phrase tables and applied stemming, POS tagging, and tweet‑specific features for subjectivity and sentiment classification. The proposed features achieved performance surpassing existing Arabic SSA results.

Abstract

Though much research has been conducted on Subjectivity and Sentiment Analysis (SSA) during the last decade, little work has focused on Arabic. In this work, we focus on SSA for both Modern Standard Arabic (MSA) news articles and dialectal Arabic microblogs from Twitter. We showcase some of the challenges associated with SSA on microblogs. We adopted a random graph walk approach to extend the Arabic SSA lexicon using ArabicEnglish phrase tables, leading to improvements for SSA on Arabic microblogs. We used different features for both subjectivity and sentiment classification including stemming, part-of-speech tagging, as well as tweet specific features. Our classification features yield results that surpass Arabic SSA results in the literature.