Abstract
With the spread of using the Internet around the world and a huge number of social media users, social media platforms have become a popular place to share customers' opinions and experiences on a variety of services and products, such as restaurants and different consumables etc. These data can be utilized to assist in decision-making. However, collecting and managing customers' opinions on different social media platforms is a complicated and challenging task. The customers express their opinions in their colloquial language that contains spelling errors and repetition of letters, so the preprocessing stage solves these problems and enhances the quality of data to become ready to conduct sentiment polarity and calculate reputation scores. In this paper, the Saudi customers' tweets are collected for four restaurants in Madinah. The Polarity Lexicon for the Saudi dialect (PLSD) and Emoji Lexicon for Sentiment Analysis (ELSA) are developed as a core component in our system to classify the extracted tweets into positive, negative, and neutral. Next, the Net Brand Reputation (NBR) is used to derive the reputation score from the lexicon feedback. Experimental evaluation demonstrates that the results from the proposed approach were compatible with the results extracted from Foursquare, an application that measures restaurants' reputations based on customer ratings and comments.