Abstract:
Sri Lanka's tourism industry generates a vast volume of tourist reviews often, providing valuable insights into traveller experiences across various attractions. However, manually analysing these reviews is time consuming and challenging due to their large volume. The research questions are, how can sentiment analysis techniques effectively classify tourist reviews, which Machine Learning (ML) models perform best in sentiment classification and how can the findings be practically implemented to enhance Sri Lanka's tourism sector. The objectives of this research are to develop a robust sentiment analysis framework for classifying tourist reviews and predict the new reviews whether positive or negative or neural, evaluate the performance of various ML and deep learning models, and provide actionable recommendations for stakeholders to improve tourist experiences and promote sustainable tourism growth. Data was collected from TripAdvisor using a web scraping tool. Sentiment labelling is done using the Vader technique to categorize reviews as positive, negative, or neutral. The review was pre-processed and converted into numerical formats using methods like TF-IDF, Bag of Words, and Word2Vec. Several ML and deep learning model including Random Forest, Naive Bayes, Decision Tree, Logistic Regression, Artificial Neural Networks (ANN), and Long Short-Term Memory Networks (LSTM) were used for classification. An ensemble approach combining Random Forest, Naive Bayes, Logistic Regression and Decision Tree achieved the highest accuracy as 89.5%, outperforming individual models. The study aimed to develop a robust sentiment analysis framework for classifying tourist reviews as objective1 and evaluate the performance of various models as objective2. The findings enable stakeholders to effectively identify key sentiment factors, address negative factors, and implement strategies to enhance the tourist experience. For instance, tourism boards can use insights to improve service quality, while businesses can tailor offerings to meet traveller preferences.
