Sentiment Analysis in Guest Reviews: An NLP Approach Applied to Portuguese Hospitality

Abstract

Guest reviews represent a valuable resource for strategic decision-making in the hospitality industry. By analysing these evaluations, managers can identify market trends, adjust management practices, reshape marketing strategies, and implement improvements in infrastructure and services. In this way, adopting a data-driven approach based on feedback directly from customers enhances the competitiveness and differentiation of hotels in the tourism market.

This study aims to evaluate not only the level of sentiment—dissatisfaction, neutrality, or satisfaction—expressed by each guest regarding a specific product or service, but also how personal and opinionated the published message is. Highly subjective texts are expected to contain emotions, opinions, and individual perceptions, whereas texts with low subjectivity tend to resemble more factual and objective descriptions.

To achieve this, a Natural Language Processing (NLP) approach was employed, involving the extraction, translation, and pre-processing of 11,810 reviews from both Portuguese and foreign visitors. The dataset is organised by hotel rating (3, 4, and 5 stars), followed by the application of two complementary sentiment analysis techniques: VADER, to determine polarity (negative, neutral, or positive), and TextBlob, to measure the degree of subjectivity.

The results show that the majority of reviews (77%) carry a strongly positive connotation, allowing the conclusion that Portuguese hospitality enjoys a high approval rate among guests. However, by identifying dissatisfaction points and satisfaction patterns specific to different customer segments, managers can better target investments and strategies to further enhance the guest experience.