Extracting place semantics from geo-folksonomies

• Main Author: Elgindy, Ehab
• Published: Cardiff University 2013
• Subjects: 004.67; QA76 Computer software
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.569953

Massive interest in geo-referencing of personal resources is evident on the web. People are collaboratively digitising maps and building place knowledge resources that document personal use and experiences in geographic places. Understanding and discovering these place semantics can potentially lead to the development of a different type of place gazetteer that holds not only standard information of place names and geographic location, but also activities practiced by people in a place and vernacular views of place characteristics. The main contributions of this research are as follows. A novel framework is proposed for the analysis of geo-folksonomies and the automatic discovery of place-related semantics. The framework is based on a model of geographic place that extends the definition of place as defined in traditional gazetteers and geospatial ontologies to include the notion of place affordance. A method of clustering place resources to overcome the inaccuracy and redundancy inherent in the geo-folksonomy structure is developed and evaluated. Reference ontologies are created and used in a tag resolution stage to discover place-related concepts of interest. Folksonomy analysis techniques are then used to create a place ontology and its component type and activity ontologies. The resulting concept ontologies are compared with an expert ontology of place type and activities and evaluated through a user questionnaire. To demonstrate the utility of the proposed framework, an application is developed to illustrate the possible enrichment of search experience by exposing the derived semantics to users of web mapping abstract applications. Finally, the value of using the discovered place semantics is also demonstrated by proposing two semantic based similarity approaches; user similarity and place similarity. The validity of the approaches was confirmed by the results of an experiment conducted on a realistic folksonomy dataset.