Producing a sentiment lexicon for colloquial Egyptian on social media: methodology and findings
Engi Amin, Mazen Hassan, Sarah Mansour · British Journal of Middle Eastern Studies · 2025
With over 3.5 billion global users on Facebook and X (formerly Twitter), social media has become a major arena for expressing sentiment, particularly in the Arab world where offline political expression is limited. This paper introduces a sentiment lexicon tailored to colloquial Egyptian Arabic, constructed from social media data. We began by scraping and preprocessing posts from Facebook and X, yielding approximately 10,000 unique words. Each word underwent two annotation tasks: polarity scoring across positive, negative, and neutral categories using a 10-point allocation system, and sentiment tagging across six basic emotions—happiness, sadness, fear, trust, disgust, and anger. Annotations were conducted by graduate student coders. For validation, we used an independent dataset of 4,000 manually annotated sentences. Evaluation results showed high accuracy across emotion categories, confirming the lexicon’s robustness. To our knowledge, this is the first sentiment lexicon for colloquial Egyptian Arabic that captures both polarity and a comprehensive range of emotional categories for data scraped from social media. Also, it is the first to rely exclusively on graduate student as sentiment annotators.