Sentiment Analysis of Yelp Review

Abstract

Digital data has developed very fast in the current era. Digital data has various forms, one of which is text data. There are a lot of text data source from many ways, such as review text data. Yelp is a local business directory and forum to review products, services, or places. We used Yelp’s review data to determine user’s sentiment or opinion about products, services, or places. Sentiment or opinion are classified into positive reviews, or negative reviews. The algorithms that used are Gaussian Naïve Bayes, Gaussian Naïve Bayes with AdaBoost, and K-NN. In this research review text data will be through a preprocessing and feature extraction stage. We used n-gram to extract the feature from the data with unigram, bigram and uni+bi-gram for indexing text. The result of this research that algorithms that has the highest accuracy rate was Gaussian Naïve Bayes with combined unigram and bigram with 86.7% from 5 fold cross-validation.