Aspect-based Sentiment Analysis on Indonesia’s Tourism Destinations Based on Google Maps User Code-Mixed Reviews (Study Case: Borobudur and Prambanan Temples)

Aspect-based Sentiment Analysis on Indonesia’s Tourism Destinations Based on Google Maps User Code-Mixed Reviews (Study Case: Borobudur and Prambanan Temples) Dian Arianto Indra Budi Faculty of Computer Science Faculty of Computer Science Universitas Indonesia Universitas Indonesia Indonesia Indonesia dian.arianto@ui.ac.id indra@cs.ui.ac.id Abstract In this paper, we conducted an aspect-based sentiment analysis using Google Maps user reviews of Indonesia’s tourism destinations which are Borobudur and Prambanan Temple. The aspects we used are Attractions, Amenities, Accessibility, Image, Price and Human Resources. The dataset obtained is in code- mixed language. We applied five machine learning algorithms which are Random Forest (RF), Naïve Bayes (NB), Logistic Regression (LR), Decision Tree (DT), and Extra Tree (ET). The evaluation performed by making eight scenarios which are the combination of stopwords removal (SR), stemming (SM), emoji processing (EP), our own stopwords dictionary (OSD), and Suciati and Budi stopwords dictionary (SSD). The model performance was measured by ten folds cross- validation. The results suggest that SM without SR, and with or without EP, SSD, and OSD did not result in a significant difference for the F1- scores. However, the combination of SM and EP, and the combination of SR, EP, and SSD did improve the performance of models for classifying sentiments. 1 Introduction Indonesia has many tourism destinations that can be visited by both domestic tourists and foreign tourists. There is an increasing number of tourists visiting Indonesia each year. Domestic tourists and foreign tourists continue to arrive every year to visit tourist attractions in Indonesia. Foreign tourists experienced a significant increase to visit Indonesia from 2009 to 2018. The average growth of foreign tourists in the 2009-2013 period was 9% per year and rose to 14% per year in the 2014-2018 period (Widowati, 2019). Indonesia in 2015 set a program of 10 Priority Tourism Destinations or called “10 New Bali” to promote Indonesian tourism and increase foreign tourist visits. One of the 10 New Bali is Borobudur Temple, Central Java. The tourism destination of temples in Indonesia besides Borobudur Temple which is quite famous for tourists is Prambanan Temple. Every year, reviews on online platforms experience significant competition. Since 2015, Google has experienced a significant increase in the number of reviews shared by internet users compared to other platforms such as Facebook, Yelp1, TripAdvisor or Foursquare (Murphy, 2018). The increase in the number of reviews on the Google platform is supported by Google’s program called Google Local Guides2. This program was originally launched in 2015 as a way to deal with Yelp Elites (Yelp contributors), which allows the most active Google Maps contributors to get rewards. In 2016, this program had 5 million contributors globally and increased every year to 120 million globally in 2019 (Sterling, 2019). In Indonesia, research on text mining in the field of tourism had been done before to obtain 1 https://www.yelp.com/ 2 https://maps.google.com/localguides important information from tourism destinations. Prameswari et al. (2017) conducted a sentiment analysis for hotels in Bali and Labuan Bajo by using five aspects of the hotel domain, those are: …

Content truncated to 500 words