Feature-based Sentiment Analysis for Slang Arabic Text

Abstract

The increased number of Arab users on microblogging services who use Arabic language to write and read has triggered several researchers to study the posted data and discover the user's opinion and feelings to support decision making. In this paper, a sentiment analysis framework is presented for slang Arabic text. A new dataset with Jordanian dialect is presented. Numerous specific Arabic features are shown with their impact on slang Arabic Tweets. The new set of features consists of lexicon, writing style, grammatical and emotional features. Several experiments are conducted to test the performance of the proposed scheme. The new proposed scheme produces better results in comparison with others. The experiments show that the system performs well without translating the tweets to English or standard Arabic.

Semantic Partitioning and Machine Learning in Sentiment Analysis

“…Considerable research is aimed at providing a complete survey of state-of-the-art works related to Arabic SA [27,[32][33][34], which employs different classification techniques in machine learning to determine the polarity of different tweets scripted in the Arabic language [4,28,35,36]. Abdulla et al [37] collected 2000 tweets on various topics, such as politics and arts.…”