International Journal on Natural Language Computing

One of the main difficulties in sentiment analysis Arabic language is presence colloquialism. In this paper, we examine effect using objective words conjunction with sentimental on classification for colloquial reviews, specifically Jordanian reviews. The reviews often include both and words, however, most existing models ignore as they are considered useless. work, created two lexicons: first includes compound phrases, while other contains associated values tendency based a particular estimation method. We used these lexicons to extract features that would be training input Support Vector Machines (SVM) classify polarity dataset have been collected manually from JEERAN website. results experiments show proposed approach improves comparison baseline models, accuracy 95.6%.