Abstract
In recent years, tourism has become a significant driver of many countries' economies. To maximize revenue from tourism, it is crucial to prioritize the effective management of scenic spots and tourist attractions, and also raise awareness about these places. Social media platforms have played a pivotal role in promoting tourism, as users frequently share videos and reviews related to tourism. Analyzing and managing these reviews is essential for understanding tourists' opinions about specific destinations. In this study, we evaluated a scenic spot by analyzing tourists' sentiments. Data was collected from popular social media sites such as TripAdvisor and Twitter using web scraping and the Twitter API. The raw data was preprocessed to remove irrelevant information and redundancies and was properly annotated for further processing. We applied two approaches to analyze the sentiments of tourists. First, we vectorized the text representing the sentiment using the term frequency-inverse document frequency (TF-IDF) and utilized big data analytics to extract meaningful insights. Secondly, we employed a pre-trained large language model, bidirectional encoder representations from transformers (BERT), with a linear classifier to classify tourists' sentiments. The results of the big data analytics approaches were compared with those of BERT and previously proposed methods. BERT outperformed other machine learning models, achieving an average accuracy of 83.5% on the test set. These insights are valuable for evaluating the informatization of tourist spots, destination management, hospitality, and overall tourist attractions.