Sentiment analysis of modern standard Arabic and Egyptian dialectal Arabic tweets

摘要

Dialectal Arabic is frequently used to express opinions on social media, but it has been scarcely researched. Most previous work on Arabic sentiment analysis systems has focused on Modern Standard Arabic, using standard verbal cues as features. This paper presents a hybrid approach for sentiment analysis of Modern Standard Arabic and Egyptian Dialectal Arabic using verbal and non-verbal cues in the form of text and emojis. Using Plutchik's Wheel of Emotions, an emoji lexicon is created as a resource for non-verbal emotion classifications. The results show, from the achieved accuracy and F-Measure of 90%, that the proposed algorithm outperforms previous work.