Sentiment Analysis Techniques for Social Media-Based Recommendation Systems

Social networks build and maintain relationships between individuals. Sentiment analysis is important in social network analysis for extracting user's interest from product preferences based on reviews to determine whether it is positive, negative or neutral review. Moreover, sentiment analysis is used to predict the sentiment of users on specific service or product received by them. In this paper, a new technique called sentiment-based rating prediction method is proposed for developing a recommendation system in which the newly introduced technique is capable of mining valuable information from social user reviews in order to predict the accurate items liked by people based on their rating. In this model, a sentiment dictionary is used to calculate the sentiments of individual users on an item. Moreover, reputations of items are computed based on the three sentiments to predict and provide accurate recommendations. In order to increase the accuracy of the outcome, then-gram methodology is added as a new feature in syntax and semantic analysis along with support vector machines for effective classification of social media data. The main advantage of the proposed model is that it considers semantics and sentiments to predict user interest and hence provides more accurate recommendations.