Service Extraction and Sentiment Analysis to Indicate Hotel Service Quality in Yogyakarta based on User Opinion

Abstract

Yogyakarta is one of the most popular tourist destinations in Indonesia. In Jogja, Malioboro is one of the most visited place for its nature, culinary, shopping, culture, and arts. Every year, in Malioboro, room occupancy rate has increased significantly. The promising tourism business had allured many online booking service provider, including Traveloka. Using Traveloka, users can reserve, give ratings and opinions on hotel services. The rating and opinion are employed to indicate hotel services quality. Later on, from the indicator, user satisfaction level can be obtained. In this study, a new approach, the rule-based method is utilized to sort out opinions containing services. Moreover, it is also employed to extract service words and opinion words. Afterward, it is utilized to identify sentiments from opinions. Because the result is still impractical, then, it is segmented according to the department and function of the hotel. From the summary, indicators of the hotel service quality can be obtained. Later, the indicators will be utilized to determine the user satisfaction level. From the experiment, it is shown that the rule-based method can be employed to sort service-based opinions and to calculate service quality indicators. The rule-based method can classify opinions based on services with precision of 0.97, recall of 1, and f-measure of 0.98. The rule-based method can classify opinions based on sentimen with precision of 0.99, recall of 1, and f-measure of 0.99. As comparison, the experiments are also performed with other methods, namely KNN, SVM, J48 and Naïve Bayes. The result shows that the rule based method achieves the best performance, compare to the other methods.