Πληροφοριακά συστήματα για την υποστήριξη λήψης επιχειρηματικών αποφάσεων στον τουρισμό
KeywordsΒαθμολογήσεις ; Ανάλυση αδόμητου περιεχομένου ; Ταξιδιωτικές κριτικές ; Τουριστική εμπειρία ; Unstructured content analysis ; Ratings ; Travel reviews ; Tourist experience
Recent advances in Information and Communication Technologies and the upgrade of the role of customers from passive receivers to active knowledge flows creators has set new demands for Tourism, as an information intensive industry, in order to adapt to the special conditions caused by the rapidly growing volume of available Big Data on the internet. It is generally acknowledged that to a large extent this data comes in unstructured form, thus demanding the adaptation of new tools and methodologies for large scale content analysis and management. Text descriptions of travel experiences in internet tourist communities are able to offer a valuable, largely unexploited, source of customer knowledge that is created spontaneously and non-intrusively. The present Thesis collects data for over 17,000 travel reviews using an automated method from the Tripadvisor tourist community including review texts, review star ratings as well as demographic and experience user data. Reviews concern three historical centers of big cities which are discrete precincts of each destination, namely Barri Gotic of Barcelona, Plaka of Athens and Trastevere of Rome. The thesis aims to develop a mixed methodological approach for unstructured content analysis of reviews by combining computerized text mining and analysis techniques with qualitative codification of concepts to capture the significant tourist experience elements regarding the three aforementioned destinations. In this context it examines, by conducting bivariate analysis as well as developing two multivariate logistic regression models, the conceptual categories derived from texts that are the most significantly associated with review star ratings, as well as the influence of user characteristics on review star ratings and tourist experience components expressed. Results include highlighting concepts categories with affective features of high valence and arousal together with the variable of historical user star ratings as the two factors most intensively associated with review star ratings. At the next stage, it is examined by what extent the analysis scheme developed can successfully distinguish the unique tourist experience attributes of each destination, by performing a comparative analysis at the bivariate analysis level and also by constructing Semantic Co-occurrence Networks of conceptual categories. This approach manages to capture in detail the major differences and the groupings of tourist experience conceptual categories for each destination.