2002, Kokla, M. and Kavouras, M., "Theories of Concepts in Resolving Semantic Heterogeneities" Print


Full title:


Kokla, M. and Kavouras, M., "Theories of Concepts in Resolving Semantic Heterogeneities", Proceedings of the 5th AGILE Conference on Geographic Information Science, 25-27 April, Palma de Mallorca, Spain, 2002. 




The need to account for the inefficient geographic information exchange between different groups of people (different domains and cultures) due to semantic heterogeneity, as well as to provide interoperability between geographic information systems, has directed Geographic Information (GI) Science towards ontological research. The role of ontology, from the philosophical, as well as from the computer science perspective, is considered important to information sharing and reuse. The development of a core geographic ontology will constitute the formalized framework for establishing detailed ontologies in more specialized fields such as environment, urban planning and social sciences and for ensuring interdisciplinary communication.

According to Sowa (2000), ontology sharing across different applications can be accomplished if these ontologies are embedded within a more general, philosophical framework. Philosophy, with its theoretical foundations and top-level categories can provide a solid framework for the integration of lower level applications (ibid.)

Our previous work, has introduced a methodology for the fusion of different geographic domain ontologies with top-level ontologies, in order to provide a solid base for information exchange and reuse. Despite the fact that a top-level ontology can indeed constitute the backbone for the integration of different domain ontologies, such an attempt reveals the diversity of existent upper-level ontologies. Therefore, it is reasonable to wonder whether the existence of
many diverse upper-level ontologies just shifts the interoperability problems to a higher level. In order to maximize usability and interoperability, GI Science needs to incorporate theories on concept formation and acquisition, as well as to take into account empirical results on human categorization. Several authors have proposed to include a cognitive aspect in the design of geographic categories.

The present paper, in order to detect the causes of semantic heterogeneity between geographic categories, explores theories of concepts, both from a philosophical and cognitive perspective. Theories of concepts and empirical studies present significant aspects of human categorization. More specifically, the paper deals with three important aspects in theories of concepts: definitions, properties and hierarchical structure. These aspects
constitute important notions, which, if incorporated in the formalization of geographic categories, could contribute to the development of more rigorous and consistent geographic ontologies.

However, since our endmost objective is the semantic and univocal understanding of geographic categories for facilitating information exchange across different applications, we attempt to incorporate these notions in the process of semantic integration. Semantic factoring is used as a tool for revealing some distinctive characteristics of geographic categorizations. Semantic factoring is the process of analyzing-decomposing the categories of the
original ontologies into a set of fundamental categories. Thus, complex concepts are analyzed into the simpler concepts out of which they are composed.

We argue that the common parts of conceptually overlapping concepts, provided that they exist, constitute the most clear, unambiguous and conceptually coherent categories. For example, to reach a consensus on different conceptualizations of the category “agricultural areas”, one has to decompose it into elementary categories: “arable land”, “permanent crops” and “pastures”.

Semantic Factoring can effectively resolve unclear cases, which correspond to variant conceptualizations across different domains. For example, CYC, WordNet and SDTS differ in the definitions and subordinates of the category “stream”. For CYC, “stream” includes both rivers and canals, whereas for WordNet only natural bodies of running water are included (rivers, branches, brooks, rivulets). On the other hand, SDTS defines a wide range
of subordinate categories for the category “watercourse”, including both natural and artificial streams of water. The common parts of these overlapping categories “stream, watercourse”, reveal the unambiguous and comprehensible basic-level categories “river” and “canal”, which reflect the consensus across the three ontologies.


The integration process using the proposed methodology helps to unclutter the integrated categorization and highlight the basic-level categories. By resolving conflicts between categories across different categorizations at whatever level they occur, we end up with the basic-level categories for all these domains (and contexts).


Download: here


[ Back ]





Do you like our new website?