Sentiment Analysis of Roman, Urdu and English Reviews

Abstract

The usage of online platforms to receive feedback, opinion or remarks of the public about a particular subject has become very common. Recommender systems can be used to predict the places users might like to visit or explore. Sentiment analysis is used to understand the latest trends, summarize the general opinion and investigate the cognitive human behavior. The aim of this project is to provide a recommendation system for users and demonstrate how sentiment analysis can be used in reviewing Roman Urdu and English reviews. We researched and analyzed the experimental results produced by different classifiers using feature selection and representation. To perform sentiment analysis, we translated a preexisting English hotel reviews dataset to Roman Urdu and analyzed the resulting corpus with the machine learning classifiers. The implication of our project is to encourage work on sentiment analysis in different languages used on the Web. The results of social analytics can assist organizations in applying the proposed methodology to the collective sentiment intelligence embedded in customers’ feedback in order to improve their product, services and marketing strategies. Our application is a recommendation system that uses sentiment analysis on Roman Urdu and English reviews to analyze the social media health of a brand.