Indonesia Tourism Sentiment Analysis Based on Social Media using Large Languange Model

Tourism in Indonesia is one of the sectors that influences the economy in Indonesia. For several years, the tourism sector has experienced a better increase in visitors. This is inseparable from digital technology that can spread information quickly and widely, such as social media. Seeing this, special attention needs to be paid to the public's response to information about tourist attractions provided to the public. Sentiment views on tourist attractions can affect the good or bad development of national tourism. The better the public response received, the more famous national tourism will be and can increase national income from the tourism sector and vice versa. The absence of a system that can analyze responder sentiment makes readers confused in drawing conclusions about the information conveyed. There needs to be a system that can analyze responder sentiment through all comments on a post so that it can be used as decision support in drawing conclusions about the information. Creating an application by utilizing one of the LLM models for classifying sentiment from each comment can overcome this problem. Fine-tuning IndoBERT and implementing it in the application of sentiment analysis of comments on social media posts of tourist attractions via Instagram and Twitter. The model was trained using a quality dataset of $\mathbf{1 5, 6 0 0}$ that had gone through the aliasing, paraphrasing, and preprocessing processes. Successfully obtained a model accuracy of 97 % in analyzing sentiment. The model is used in applications that are used to support public decisions regarding sentiment assessments of tourist attractions that are influenced by social media posts.