Sentiment analysis of customer reviews: Accommodation establishments

Accommodation businesses in Kemer and Chania, which are close competitors to each other, endeavour to get a bigger share in the market by satisfying their customers with the goods and services they offer. For this purpose, accommodation businesses prepare strategic plans by making competitor analyses. How the goods and services offered are perceived by the customer and their emotional state shows how much the objectives are realised. Customer comments on TripAdvisor website are considered to be an important data source for the evaluation of accommodation businesses. Based on this idea, customer reviews of accommodation establishments in two regions between 2015 and 2022 were excavated, and a total of over 197 thousand data were obtained. As a result of various normalisation processes and screening, more than 173 thousand comments were divided into 7 categories (staff, room comfort, fees, cleaning, recreation, food and beverage, hotel general) and 26 sub-dimensions. The data were analysed using machine learning, quantitative and qualitative data analysis techniques. Many methods were tried in machine learning and the KNN method (75%) provided the highest success. In the multiple linear regression analyses, the ratio of the emotion scores obtained from the categories and sub-dimensions to explain the score given by the customer was found to be 37.6% on category basis and 37.9% on sub-dimension basis in average. The data were also compared according to destination, season and year variables. In the recreation category and all sub-dimensions and only in the dessert, fruit and snack buffet sub-dimensions of the food and beverage category, it was found that the emotion scores of the customers staying in Kemer region were higher. In all other categories and sub-dimensions, it was determined that the emotion scores of the customers staying in Chania region were high. In the analyses made according to the season variable, it was determined that the emotion scores of the customers in the low season were significantly higher in general. In the analyses made according to the year variable, it was determined that the average customer emotion scores have increased in most categories and sub-dimensions from past to present. However, it was determined that in the categories and sub-dimensions of food and beverage, personnel, room comfort, the average emotion score decreased during the pandemic period, increased in 2022 and reached 2019 levels. The linguistic expressions of customers according to the variables in all sub-dimensions were analysed in depth with NVivo and MAXQDA programs and the results were presented factually.