A lightweight ontology for landmarks to assist rescue in mountainous areas

Abstract. When people are injured or lost in mountains during outdoor activities and when web-based locations are not available, they locate themselves by describing their environment, routes and activities. The description of their location is done using landmarks and spatial locations (e.g., “I am located in front of Punay Lake”, “I am near a protected area”). Landmarks used can be named (e.g., “Punay Lake”) or unnamed if the landmark has no name or if the victim does not know it (e.g., "area lake"). Landmarks are represented in geographic databases by name (if possible), type and geometry. To reduce the heterogeneity of landmark types present in oral language and geographic databases representing landmarks, and thus improve locating victims, our goal is to define a controlled vocabulary for landmarks. In this research, we present a lightweight ontology (i.e. ontology having generally less complexity and does not express formal constraints) of landmarks, named Landmark Ontology (OOR), describing landmark types. It is an application ontology, i.e. it is designed to support mountain rescue operations. The ontology construction is adapted from the SAMOD methodology for engineering ontology development and involves researchers and experts from mountain rescue teams. The construction of OOR is composed of four main phases: knowledge acquisition, conceptual formalisation, implementation, and testing. The implementation phase is carried out by an iterative and collaborative approach and using four formalised sources of knowledge (a landmark and a landform ontologies, and two other domain vocabularies), an un-formalised taxonomy of outdoor activities, and five authoritative and volunteered geographic information sources representing geographic data. The landmarks ontology contains 543 classes associated with 1739 labels: 1086 prefLabel (preferential label) in French and English, 321 altLabel (alternative label) in French, and 332 altLabel in English. The depth of the ontology varies from four for land cover, hydrological and land subdivision landmark types), to six for landform types, and eight for building types. Although the use of ontology is broader, in this paper we illustrate and test its use through three applications in the context of mountain rescue operations: semantic mapping, data instantiation and data matching.