Along with an advent of the 4th industrial revolution, paradigms in various domestic industries have changed, where in the field of ‘hospitality’ is experiencing diverse changes based on digital technologies comprising big data and artificial inte...
Along with an advent of the 4th industrial revolution, paradigms in various domestic industries have changed, where in the field of ‘hospitality’ is experiencing diverse changes based on digital technologies comprising big data and artificial intelligence etc. In particular, consumers are writing and sharing their experiences of services they had received real-time in accordance with increasing access to internet via the popularization of smartphones by which the data thereof are increasing explosively. Such a vast amount of review data can play the important role over entire domain of consumers’ decision making, as well as the roles of delivery of information and recommenders wherein the positive reviews on services they had experienced increase the loyalty of consumers. In addition, the review data reflect diverse kinds of consumers’ emotions frankly expressed thus they can be exploited as resources to measure and improve consumers’ service experiences.
A survey employing questionnaire is widely used to measure customers’ experiences through which diverse kinds of questions can be used for quantification of consumers’ experiences however it has disadvantages of measurement errors dependent on limitations of sampling and errors in memory of past experiences of respondents, terminologies in questions, and length of questions etc. On the contrary, the analysis using text mining techniques has an advantage of identifying psychological aspects of respondents realized by the broad collection of all written opinions in categories freely set by consumers which was difficult to identify through conducting quantitative studies despite disadvantages of selecting appropriate respondents and difficulty in measurement according to intended designs of the analysis.
The present study intends to read the emotional factors of satisfaction and dissatisfaction of foreign tourists by using the left on-line reviews of restaurants, and to verify the presence of positive effects of sentiment polarity values of selection attributes upon overall level of customers’ satisfaction and differences in types of restaurants. To conduct the study, the data of on-line reviews on hotels and restaurants of customers were collected from the Review Community (TripAdvisor). The crawler was prepared by using Python to collect the data from corresponding community by which the profile information of customers, reviews written in English, and comprehensive scores thereof were collected. And then, the text mining techniques were exploited for the analysis of each text. The natural language processing technique was used to quantify the text data, and the frequency analysis on keywords per each type of restaurants was carried out.
And then, the topic modeling technique was used to distinguish each sentence in reviews in terms of selection attributes of the foods, price, service, and atmosphere, and then the distinguished sentences were put into the sentiment analysis to determine the factors of satisfaction or dissatisfaction of tourists. Finally, the test of hypotheses was carried out by using techniques of statistical analysis on the effectual relationship between differences in sentiment polarity at each type of restaurants and degree of satisfaction of customers.
The results obtained from the present study are summarized as in the following.
First, the factors of satisfaction and dissatisfaction per each type of restaurants were compared to each other through conducting sentiment analysis on the on-line reviews left by tourists. Common terms used to express dissatisfaction of tourists comprised time, staff, wait, late, forget, missing, and wrong etc. The terms imply the degree of hospitality of employees in restaurants and management of waiting time of customers in restaurants could influence greatly on the degree of dissatisfaction of customers. In respect of each type of restaurants, the terms comprising expensive, quality, course, table, reservation, credit, card, glass, plate, table, lobby, cat, and taxi etc. appeared in fine dining restaurants, whereas the terms such as English, speak, language, queue, and busy etc. appeared in casual dining restaurants. The terms such as line, quick, fast, seating, parking, wrap, and dirty were frequently used in describing the quick and self-service restaurants. This suggests the factors of dissatisfaction vary according to each type of restaurants.
Second, the demographic characteristics of customers were compared by using the results of sentiment analysis. Contrary to customers visiting restaurants who left many reviews on respective services and foods, the reviews on prices appeared relatively few, wherein the difference in scores of emotion on selection attributes between sex and age of customers appeared insignificant. Besides, the ratio of dissatisfaction on price and service appeared higher than that of foods and atmosphere of restaurants from the analysis of customers divided into groups of the satisfaction (positive) and dissatisfaction (negative) of selection attributes. This suggests the service of restaurants requires continuous awareness and management in that the failure in providing pertinent service could lead to the dissatisfaction of customers.
Third, the differences in values of sentiment polarity of selection attributes per each type of restaurants were compared. The value of sentiment polarity of selection attributes of foods, service, and atmosphere of fine dining restaurant appeared highest, whereas the highest value of sentiment polarity of selection attribute of price appeared in the quick and self-service restaurant. In respect of the appraisal of tourists on the high-class fine dining restaurants, they appeared satisfactory on foods, service, and atmosphere therein while they exhibited dissatisfaction on prices of restaurants.
Fourth, the ANOVA tests per each type of restaurants were carried out to determine the presence of significant differences in values of sentiment polarity of selection attributes. The results revealed the presence of significant differences of all selection attributes per each type of restaurants.
Fifth, the effect of sentiment polarity of selection attributes per each type of restaurants upon degree of satisfaction of customers was examined. Values of sentiment polarity of foods, price, service, and atmosphere of restaurants upon customers’ satisfaction were verified by conducting the Tobit regression analysis. The values of sentiment polarity of food, price, and service appeared with significant effect on customer’s satisfaction, whereas the effect of the value of sentiment polarity of atmosphere of restaurants appeared insignificant. This was concluded the preference to foods, price, and service than atmosphere of restaurants of practical tourists affected relatively more on the degrees of satisfaction.
In the present study, the factors of satisfaction and dissatisfaction of tourists reflected in on-line reviews were analyzed and compared to each other by using the sentiment analysis of selection attributes per each type of restaurants based on the topic modeling technique, by which the study demonstrated the quantification of customers’ experiences was enabled through the information of text. Together with ordinary approaches that measured customers’ experiences by exploiting conventional surveys employing respective questionnaires, the study approach to text review also demonstrated that it could bring significant research results.
Keywords
selection attribute of restaurant, on-line review, big data, text mining, natural language processing, topic modeling, sentiment analysis