Abstract

People love experimenting on their food with different tastes. And when it’s about visiting the restaurant or ordering food online, they will definitely look for the reviews which will talk about the aspects like services, ambience and cost along with food quality. This food problem is not a single day problem, it’s getting repeated every day. Some end up with positive reviews and few end up with negative or neutral reviews. In this work a framework is developed called ‘RestoQ’, which uses text analytics for sentiment analysis at the aspect level to discover and rank the restaurants. The framework analyzes the reviews for the sentiments across four aspects – price, food quality, service quality and ambience. Unsupervised lexicon-based classifier and a naïve Bayesian classifier are used to evaluate and score the sentiments at aspect level. The final score will be a combined sum of each score for the review, which requires further work rank the aspects based on reviews. Surprisingly unsupervised method out performs the supervised method. It is proposed to extend the work with context based methods using word2vect and LSTM.

Published in:

7th International Conference on Business Analytics and Intelligence 5-7 December 2019 IIMB.