Opinion mining On Indonesian Tourism TikTok Video Content using FastText and Multilayer Long Short-Term Memory
Dony Ariyus1, a) and Daniel Manongga2, b, Irwan Sembiring3, c)

1 Faculty of Computer Science, Universitas Amikom Yogyakarta, Yogyakarta 55281, Indonesia
*dony.a[at]amikom.ac.id
2,3 Fakultas Teknologi Informasi, Universitas Kristen Satya Wacana, Salatiga 50711, Indonesia
b)danny.manongga[at]uksw.ed, c)irwan[at]staff.uksw.edu

Abstract

Analysis of social media is a topic of current research discussion. The emergence of web 2.0 technology is the primary reason that has transformed social media into a digital platform that facilitates the expression and sharing of opinions on various topics through diverse content. Expressions and opinions that emerge through interactions between social media users have the potential to be investigated and utilized in a variety of contexts, including by the government, in order to comprehend the thoughts of its citizens regarding newly implemented public policies. Currently, the government of the Republic of Indonesia, via the ministry of tourism and creative economy (Kemenparekraf), has established regulations for five super priority tourist destinations: Lake Toba, Labuan Bajo, Borobudur, Mandalika, and Likupang. The policy^s success needs to be analyzed based on opinion mining to determine the number of citizens who recommend the destination. As one of the most popular social media platforms available on the market today, TikTok has the potential to be used to explore opinions regarding selected tourist destinations. This is because many young tourism performers utilize TikTok to express their thoughts and feelings. Opinion Mining on TikTok social media data has its challenges because the use of language is not standard and is mainly done using slang as interactions are carried out daily. The use of language corpus, widely circulated, gives poor results in analyzing public sentiment. To investigate public opinion regarding Tiktok social media data, FastText and single, double, and triple layers of FastText and LSTM were employed in this study. Consequently, the employment of FastText and LSTM with multiple layers delivers good performance and has the potential to be employed in a variety of system innovations to investigate public opinion, particularly on TikTok social media data.

Keywords: Social Media Analysis, TikTok, Opinion Mining, FastText, Multilayer LSTM

Topic: Big Data and Analytics