Aspect Analysis of Cebu Establishments' Online Reviews using k-means Clustering and word2vec

Abstract

Customer reviews are important part to any business. With the development of the technology, customer reviews are usually found on the internet. In this study, online reviews from different Cebu establishments were gathered using selenium as web scraper. A total of 3776 online reviews were gathered. Word2vec and k-means clustering were utilized to analyze and discover different online review corpora. To identify the best number of clusters, a series of experiments were conducted to find for the best Silhouette coefficient. For better analysis of k-means clustering, open coding was used to understand the significant qualitative codes. Based on the k-means clustering results, the following qualitative codes were identified: time, staff, friendly, service, affordable, love, food, price, ambiance, good, great, relax. Analyses of the clusters show that quality service, tasty and affordable food and good atmosphere are the significant aspect that the online reviews are concerned. Based on the word2vec results, the researchers focused on the following words: Waiters, relax, great, ambiance, service and tasty. The results of the study provide meaningful insights on the group of words obtained using the analogy to word2vec model, as well as the subject focus of the categories.