Mid Sweden University

Although globally Google Maps ranges among the most popular web-portals for travel information search, tourism studies using Google Maps data are scant. We are reporting about ongoing research conducted in Sweden aiming both, at reliably monitoring the fragmented cultural tourism offer as well as analyzing travelers’ complex cultural tourism experience by using Google Maps data. More precisely, the supply-side monitoring estimates and visualizes Google places labelled by suppliers as ‘cultural’, thereby revealing regional patterns and the geographical distribution of the cultural tourism offer all over Sweden. By contrast, in order to reveal travelers’ experience outcomes, demand-side analytics focus on user generated content [UGC] analysis by applying sentiment analysis and topic detection, respectively. Before discussing the findings, we briefly outline the used methods for data extraction: First, a grid-field for Sweden created with the geographical information system ArcGIS served as input for the Google Maps places API to retrieve 115,316 Google places. Subsequently, a sub-total of 13,915 place types relevant for cultural tourism was identified through manual annotation. Finally, place types were mapped to cultural tourism place categories as proposed by the literature, such as heritage, arts, religious, and natural heritage tourism, respectively. With regard to UGC analytics, the web-crawler Scrapy was employed to extract 353,960 tuples with review-text related to cultural tourism places. While for sentiment analysis a lexicon-based approach using Liu’s (2020) famous wordlists for positive and negative tonality was employed, Latent Dirichlet Allocation (LDC) was applied to deduce topic-clusters that best represent cultural tourism categories. As a final step, retrieved topic-clusters were mapped to place types enriched by UGC-based sentiment. By using the visualization software Tableau, findings show the share and geographical distribution of cultural tourism place categories for Sweden (national view) as well as for cultural tourism place types for each Swedish region (regional view). Moreover, most popular as well as top-rated topic-clusters along with most frequent opinion words can be displayed for each cultural tourism place type. Most notably, sentiment distribution (i.e. positive, negative, and neutral) can be shown for place categories for each region and over time. We conclude that most relevant analysis perspectives for real-time tourism monitoring and analytics are adequately supported by the inexpensive Google Maps data. As limitation, we point at potential representativeness issues, as 56% of data sets do not comprise any review. For future research, we envisage to also include adjectives from UGC data for better grasping travelers’ complex cultural tourism experience. Finally, we propose the analysis of traveler’s spatial behavior and movement patterns by employing association rule analysis and sequential pattern mining, respectively.