Aspect Based Sentiment Analysis of Restaurant Reviews using Ensemble Algorithms

Abstract

Digital information is continuously generated from various sources such as social media, and user reviews for services. Analysing of these reviews to extract user opinions is critical for developing customer satisfaction. In particular, restaurant services may be improved with customer feedback if the user opinions or sentiments are inferred from user reviews. The development of automated software systems for evaluating customer reviews is ongoing. This research work proposes an ensemble machine learning approach on the basis of the lexicon method and machine learning classifiers and the performance of the new model is evaluated on the basis of accuracy.  The Nave Bayes (NB), Decision Tree, and Support Vector Machine (SVM) are used to create an ensemble Classifier. The article presents a comparative assessment of the efficacy of ensemble methods for sentiment analysis. The restaurant review dataset is used to demonstrate the feasibility and benefits of the proposed approach.  The work focuses on not only the overall opinions but also aspect-based opinions, including Food, Service, Ambience, Quality and Price. The analysis is followed by a wide range of comparative experiments demonstrating the efficacy of ensemble technology for sentiment analysis. Finally, some detailed discussion and conclusions are drawn concerning the effectiveness of ensemble technology.