Semantic similarity analysis of user-generated content for theme-based route planning

Several technical approaches to a touristic tour planning, which connect popular points and routes of interest or provide locations related to specific themes, have been published in recent years. Hereby, points of interest are found and evaluated on the basis of user-generated web content. However, no approach exists to the author's knowledge, which allows truly individual theme route planning. Individual means, that a user flexibly defines start point and destination and receives an optimised route, which will guide him through a townscape/landscape with most interesting features being situated along the proposed way. We introduce two methods to find such an individual theme route based on user-generated content. The basis for both methods is the determination of semantic similarity between a selected Wikipedia concept (e.g. a specific architectural style) and other geo-referenced Wikipedia concepts (e.g. a building). The first method has been termed the continuum method: it uses semantic similarity measures together with a density distribution from theme-related, geo-tagged photos in the web, in order to create a continuous ‘surface of attractiveness’. Such a conceptual continuum can – together with the static geometric length of network features – form the basis of an assignment of impedance values to a navigation graph. The second method has been termed the spot sequence method: it models the theme route as a specific version of the travelling salesman problem. A route is composed by sequentially adding visit points to a navigation graph from the start to the end point. Priorities are derived from the ranked semantic similarity values. The achieved results have been compared and evaluated on a basis of a user survey.