Domain-Level Topic Detection Approach for Improving Sentiment Analysis in Arabic Content

Abstract

Social networks are considered today as the most popular interactive media where people can communicate, share information and express opinions without any limitation. The interest of the scientific community towards social contents has increased due to their importance in various fields such as marketing, sociology and politics. Several research areas related to social networks have emerged namely, community detection, sentiment analysis and topic detection. In this paper, we propose a domain-level topic detection approach for improving sentiment analysis in Arabic social content. The proposed approach is based on a supervised learning technique on Arabic collected data. Training dataset is mainly composed of Arabic press articles, while the test dataset is represented by posts and comments extracted from Arabic Facebook pages. Experimental evaluation showed that the proposed approach achieves good performances with precision values between 75.36% and 97.89%.