From reviews to emotions: Analysing Bragança’s tourism attractions on TripAdvisor

Abstract Over the past decade, sentiment analysis has emerged as a pivotal tool in tourism-related texts, driven by the sheer volume of tourist attractions and the wealth of online information. Tourists increasingly turn to travel websites to access specific information that often eludes standard evaluations of tourist attractions. Forums particularly illuminate specific information needs and their ties to potential destinations. Among these platforms, TripAdvisor has become a favoured choice for posting reviews, ratings, and facilitating online bookings. In this context, this study aims to analyse and assess sentiment in reviews sourced from the online platform TripAdvisor, focusing on tourist attractions in the northern Portuguese destination of Bragança. The research spotlights the disparity between qualitative and quantitative rankings. The study also underscores the importance of data pre-processing, including removing irrelevant information and stop words. Pre-processing was crucial in refining sentiment prediction accuracy, highlighting the differentiated roles of these words in context and meaning. Despite utilising advanced techniques such as tokenisation, TF-IDF weighting, logistic regression, and n-grams, the study’s models encountered challenges in achieving high accuracy in sentiment prediction. Even the incorporation of bigrams did not yield substantial performance improvements, with the models frequently inclined to overestimate negative and positive sentiments.