Loading...
Publication
Monitorização da reputação gastronómica através das Online Travel Reviews: uma abordagem com LLMs e LDA em restaurantes Michelin do Grande Porto
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 1.91 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A presente dissertação analisa as experiências gastronómicas relatadas por utilizadores em Online Travel Reviews (OTRs) de restaurantes com estrela Michelin localizados na região do Grande Porto. O estudo insere-se num contexto de crescente valorização da gastronomia como fator diferenciador de destinos turísticos, sendo as experiências alimentares reconhecidas como um dos principais motores da satisfação, fidelização e recomendação dos visitantes. No caso dos restaurantes de alta gastronomia, o impacto da reputação digital torna-se ainda mais expressivo, influenciando diretamente a perceção de valor e a decisão de consumo. Apesar da crescente centralidade da temática, existe ainda escassa investigação empírica que explore, com base em grandes volumes de dados textuais, a forma como os clientes descrevem essas experiências. As OTRs surgem, assim, como uma fonte rica e espontânea de informação, permitindo aceder à voz dos consumidores em contexto real.
Tendo em conta esta pertinência, recorreu-se a uma metodologia inovadora que combina técnicas de análise não supervisionada com modelos de linguagem de grande escala (LLMs), baseada num estudo atual que identifica cinco dimensões-chave de experiências gastronómicas memoráveis: sensorial, afetiva, intelectual, comportamental e social. Os dados da investigação são constituídos por avaliações em inglês, publicadas na plataforma digital TripAdvisor entre janeiro de 2020 e maio de 2025. A análise baseou se em três técnicas principais: modelação de tópicos com Latent Dirichlet Allocation (LDA), análise de clusters com o algoritmo K-means e análise de associações de termos.
Os resultados revelam que as dimensões sensorial e afetiva são as mais frequentemente mencionadas, com destaque para expressões como "delicious", "memorable", "fantastic" e "presentation". Através da segmentação, foi possível identificar perfis distintos de utilizadores, evidenciando a diversidade de perspetivas na avaliação da experiência gastronómica. A análise de associações revelou ainda como termos afetivos, sensoriais e comportamentais se interligam, contribuindo para a construção de uma experiência rica e significativa.
Este trabalho oferece contributos práticos e teóricos relevantes, ao propor uma metodologia replicável para a monitorização da reputação online e para o desenvolvimento de estratégias de fidelização e diferenciação no setor da restauração de excelência. Sugere-se ainda o alargamento futuro da análise a outras regiões, idiomas e plataformas digitais, bem como a aplicação de modelos supervisionados que permitam prever o impacto de determinadas dimensões na satisfação global dos clientes.
This dissertation analyses gastronomic experiences reported by users in Online Travel Reviews (OTRs) of Michelin-starred restaurants located in the Greater Porto region. The study is set against a backdrop of increasing recognition of gastronomy as a key differentiating factor for tourist destinations, with food experiences now acknowledged as major drivers of visitor satisfaction, loyalty, and recommendation. In the case of fine dining establishments, digital reputation has a particularly strong impact, directly influencing perceived value and consumption decisions. Despite the growing importance of the topic, empirical research exploring how clients describe these experiences based on large-scale textual data remains scarce. OTRs thus represent a rich and spontaneous source of insight, allowing access to the consumer voice in real-world contexts. Considering this relevance, an innovative methodology was adopted, combining unsupervised analysis techniques with large language models (LLMs), based on a recent study that identifies five key dimensions of memorable gastronomic experiences: sensory, affective, intellectual, behavioural, and social. The dataset comprises English language reviews published on the TripAdvisor platform between January 2020 and May 2025. The analysis was structured around three core techniques: topic modelling with Latent Dirichlet Allocation (LDA), cluster analysis using the K-means algorithm, and term association mapping. Results show that the sensory and affective dimensions are the most frequently mentioned, with notable expressions such as "delicious", "memorable", "fantastic", and "presentation". Segmentation allowed for the identification of distinct user profiles, highlighting diverse perspectives in the evaluation of dining experiences. Association analysis further revealed how affective, sensory, and behavioural terms are interlinked, contributing to the construction of rich and meaningful experiences. This work offers relevant practical and theoretical contributions by proposing a replicable methodology for online reputation monitoring and the development of loyalty and differentiation strategies in the fine dining sector. Future research is encouraged to extend the analysis to other regions, languages, and digital platforms, as well as to explore supervised models that may help predict the impact of specific experience dimensions on overall customer satisfaction.
This dissertation analyses gastronomic experiences reported by users in Online Travel Reviews (OTRs) of Michelin-starred restaurants located in the Greater Porto region. The study is set against a backdrop of increasing recognition of gastronomy as a key differentiating factor for tourist destinations, with food experiences now acknowledged as major drivers of visitor satisfaction, loyalty, and recommendation. In the case of fine dining establishments, digital reputation has a particularly strong impact, directly influencing perceived value and consumption decisions. Despite the growing importance of the topic, empirical research exploring how clients describe these experiences based on large-scale textual data remains scarce. OTRs thus represent a rich and spontaneous source of insight, allowing access to the consumer voice in real-world contexts. Considering this relevance, an innovative methodology was adopted, combining unsupervised analysis techniques with large language models (LLMs), based on a recent study that identifies five key dimensions of memorable gastronomic experiences: sensory, affective, intellectual, behavioural, and social. The dataset comprises English language reviews published on the TripAdvisor platform between January 2020 and May 2025. The analysis was structured around three core techniques: topic modelling with Latent Dirichlet Allocation (LDA), cluster analysis using the K-means algorithm, and term association mapping. Results show that the sensory and affective dimensions are the most frequently mentioned, with notable expressions such as "delicious", "memorable", "fantastic", and "presentation". Segmentation allowed for the identification of distinct user profiles, highlighting diverse perspectives in the evaluation of dining experiences. Association analysis further revealed how affective, sensory, and behavioural terms are interlinked, contributing to the construction of rich and meaningful experiences. This work offers relevant practical and theoretical contributions by proposing a replicable methodology for online reputation monitoring and the development of loyalty and differentiation strategies in the fine dining sector. Future research is encouraged to extend the analysis to other regions, languages, and digital platforms, as well as to explore supervised models that may help predict the impact of specific experience dimensions on overall customer satisfaction.
Description
Dissertação de mestrado
Keywords
Experiências gastronómicas Online Travel Reviews (OTRs) Machine learning TripAdvisor Reputação online Gastronomic experiences Online reputation