Feature comparison of yelp restaurants based on sentiment analysis

Abstract

Today the internet has given a huge space for people to explore new experiences. It has made lives more comfortable. Global information can be retrieved within seconds over any topic. Advance of such internet technologies made people to share own experiences over various topics. The extensive use of social media, forums, blogs or various e- commerce platforms have made this easier. People give opinions on various products or services based on their experiences. These reviews affect marketing strategies of online businesses. Also, comments are important for the staff or the owners as well. Positivity or the negativity of these comments is important. But thousands of reviews are collected per a day. In order to get an idea about the reviews one has to spend many hours reading them. This is an impossible task. Hence there should be an easier way to explore huge number of reviews with in seconds. This research is based on implementing a system to facilitate this task. The domain of the business is restaurants. People often search for good restaurants. Most of them are used to explore various customer reviews to find the best one. The solution will be a hybrid sentiment-based system which aids customers to find good restaurants in a particular city. Customer review datasets of various restaurants in a particular city are collected from Yelp.com. The reviews are analyzed based on popular restaurant features such as price, quality of food, ambience or service. The main functionality of the system is to deliver a sentimental comparison between various restaurants over their features. Several NLP tasks with machine learning techniques such as multilabel Naïve Bayes model, SVM model, deep learning convolution neural networks and word embeddings are used in this research.