An Enhanced Approach for Arabic Sentiment Analysis

Abstract

In recent years, there are massive numbers of users who share their contents over wide range of social\nnetworks. Thus, a huge volume of electronic data is available on the Internet containing the users’ thoughts,\nattitudes, views and opinions towards certain products, events, news or any interesting topics. Therefore,\nsentiment analysis becomes a desirable topic in order to automate the process of extracting the user’s\nopinions. One of the widely content sharing languages over the social network is Arabic Language.\nHowever Arabic language has several obstacles that make the sentiment analysis a challenging problem.\nMost users share their contents in informal Arabic. Additionally, there are lots of different Arabic dialects.\nHence, Arabic sentiment analysis researches is developed slowly compared to other languages such as\nEnglish. This paper proposes a new hybrid lexicon approach for Arabic sentiment analysis that combines\nin the same framework both unsupervised and supervised technique. In the unsupervised phase, the polarity\nof data is extracted by means of Look-up table stemming technique. In the supervised phase, we use the\ndata of the true classified polarity from the unsupervised phase to generate and train a classifier for the\nfurther classification of the unclassified data. We test and evaluate the proposed approach using MIKA\ncorpus [1]. The results show that the proposed approach gives better results.