Characterising sentiment spill-over in restaurant reviews

Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Knowledge Management and Business IntelligenceUsual restaurants’ attributes such as menu price, cuisine type, and overall rating are key factors often considered by consumers when deciding to go out to eat. Such metrics are also largely explored by virtual communities, e.g. TripAdvisor, which allow users to rate restaurants and share their experience through online reviews. Such model raises the question whether sentiment spill-over occurs based on these features or, in other words, whether specific restaurant characteristics can influence or bias sentiment orientation. Using Network Analysis, online consumer reviews, and lexicon-based Sentiment Analysis this study presents an approach to identify whether restaurants with similar characteristics share similar sentiment orientation. Although results showed that it is possible to map sentiment orientation to restaurant characteristics, no evidence was found of a deterministic proxy between these features with sentiment orientation, meaning that the sentiment towards a restaurant is likely to be influenced by other factors rather than menu price, cuisine type or rating. In addition, the analysis showed that restaurants in Lisbon have an overall good rate and that the rates correlate positively with favourable sentiments for both Portuguese and English languages