Sentiment Analysis for Standard and Dialectal Arabic Using the Lexicon-Based Approach

Sentiment analysis offers a substantial opportunity to extract valuable information by studying trends and discovering of new insights. However, conducting such an analysis based on opinions expressed in Arabic proves to be challenging task due to the nature of this language and the strong presence of dialectal expressions. This work focuses on sentiment analysis of Arabic opinions using the lexicon approach. Thus, we propose a new sentiment analysis lexicon that covers modern standard Arabic and Moroccan dialect. First, we present the construction process of this lexicon according to a manual step and automatic expansion step. Then, we will go through the steps of implementing it in order to classify sentiments. We evaluate the performance of the lexicon through two Arabic corpora collected from Facebook and Twitter. The results obtained showed good scores, since the classification performance was above 80% in the Fscore for both corpora. When it comes to lexicon coverage, the score was above 90% for the same tested corpora.