Modelling Personalised Landmarks
Many websites no longer show each user the same content. The pages are filtered by personal preferences based on collected behavioral information of the user. Although mobile routing services for pedestrians are widespread available and used by many people in everyday live the concept of personalisation has not yet been implemented for this domain. Studies from cognitive psychology have shown that landmarks are crucial components of route directions. The landmarkness of an object is dependent on the so called landmark dimensions dependent on the object’s visual, semantic, and structural characteristics. Currently available conventional landmark identification models consider only these landmark dimensions, which are static and dependent on an object itself. However, whether an object becomes a landmark is not only affected by the characteristics of an object itself but also by the perspective of the traveller. This means the landmarkness of an object depends on personal dimensions of the traveller such as: prior spatial knowledge, personal interests, personal goals, personal background, and individual traits, but also on other not yet identified dimensions.
We assume that the data collection for these personal dimensions is the highest effort for the identification of personalised landmarks for route directions. Within this project we develop models for personalised landmark identification considering landmark and personal dimensions and compare their results with conventional, non-personalised models considering only landmark dimensions. We collect data for landmark and personal dimensions of objects along an innercity route through Augsburg to feed the landmark identification models. The landmark dimensions are extracted from official databases or acquired during field surveys. We collect personal dimensions with a survey using ESRIs Survey123. The survey contains questions about the personal background, the personal interests, and the prior spatial knowledge at the decision points of the route. Furthermore, we ask survey participants to select one object they like as a landmark and one object they do not like for a personalised route direction.
Then, we compare the results of the personalised landmark identification models with the results of the conventional, non-personalised landmark identification models. We perform analyses to find out whether there are statistically significant differences between the conventional and the personalised models. As a result of the project we expect an answer to the question, whether the additional data acquisition effort for personal dimensions is worthwhile or whether it is sufficient to focus on conventional, non-personalised models and their integration in existing mobile routing services.