Implicit interaction profiling for recommending spatial content

  • Authors:
  • Joe Weakliam;Michela Bertolotto;David Wilson

  • Affiliations:
  • University College Dublin;University College Dublin;University of North Carolina at Charlotte, University City Blvd, NC

  • Venue:
  • Proceedings of the 13th annual ACM international workshop on Geographic information systems
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

When individuals request task-relevant spatial content in the form of area maps, GIS applications typically return default maps displaying standard map content. Little effort is made by these applications to present users with personalized maps displaying spatial content tailored to users' specific interests. Maps generated usually contain superfluous information that hinders the user's end goal and is irrelevant in terms of their spatial content preferences. Users may then customize the map through toggling features on and off but this must be done repeatedly whenever they request a map. One solution is to demand explicit input from users, before presenting them with a map, detailing features of interest related to their current task. This, however, proves an expensive answer as the system is reliant on user input. Another solution is to store simplistic profile information whereby the user ticks several feature boxes. While simple customizations could be stored, only basic interaction information is captured in the user profiles. We outline an approach to solving this problem by providing personalized maps whereby only the most relevant spatial content is returned each time a user requests a map. Map personalization is realized by monitoring users' implicit interactions with maps when locating content and regions of interest. User preferences regarding map features and zones of interest are inferred from the actions executed. This is an attractive solution, as it requires no real effort from the user, other than standard usage. All map interactions are captured at the interface and the system learns users' interests by unobtrusively observing their behavior. A persistent user model storing information describing user interests related to spatial content is created and evolves over time.