Abstract
Sentiment analysis (SA) is a modern text mining disciplinary that gained notable position due its various application in social networks (SN) and many internet domains. Since, it is used for discovering audience directions, and impressions about products or any subjects discussed in the internet via social media. Personal opinions availability in SN gave SA a significant attention to discussions makers on modern corporation. In this paper, we uses machine learning techniques via many features and same corpus states in Arabic language, comparing their results, and illustrating the significance of terms merging and pruning in various figures, which helps in the field of SA performance increasing purposes.