Semantic Similarity based on Weighted Ontology - Geographical Information Systems: Trends and Technologies - page 294

The model of Tversky has also been used in the context of ontologies
(Pinto and Martins 2004) in (Kavouras et al. 2005) to determine semantic
similarity of geographic categories (or entities). In particular, the problem
of cross-mapping of geographic ontologies has been addressed and a
systematic methodology for comparing categories has been presented.
Essentially, this approach is addressed to establish semantic similarity
among categories from three geographic ontologies (i.e., CORINE LC, 5
MEGRIN 6 and WordNet) by using information derived from categories'
defi nitions. More specifi cally, the focus is to extract semantic properties-
relations and values for measuring the similarity of categories. For instance,
the category “lake” has been defi ned as “a body of water surrounded by
land” in MEGRIN, while the same category in WordNet is defi ned as “a body
of (usually fresh) water surrounded by land”. In both cases, the semantic
properties and relations that can be identifi ed are as follows: Hypernym with
value “body”, Material with value “water” and “water (usually fresh)”,
respectively, and Surrounded-by with value “land”. Obviously, the above
mentioned ontologies equivalently defi ne the category “lake”. Similarly,
“Ditch” in WordNet is defi ned as “any small natural waterway”, while
in MEGRIN it is defi ned as “A canal for irrigation and drainage”. These
categories share only Hypernym with value “waterway” and “canal” (in
Fig. 1, Canal is a generalized concept of Waterway ), while “irrigation and
drainage” can be identifi ed as Purpose in MEGRIN, and “small”, “Natural”
as Size and Nature, respectively in WordNet. Thus their similarity measure,
according to Tversky's model (see above for α = β = 1) is equal to 0.25.
In Schwering and Raubal (2005), the authors propose a method based
on spatial relations between different geospatial concepts. To capture the
semantics of geo-objects with spatial relations the authors use a reduced
group of spatial relations extracted from the set of spatial relations in
natural language formalized by Shariff et al. (1998). In order to investigate
similarities, they consider hydrological geo-objects within a large-scale
topographic map for defi ning spatial relations. As a case study, they consider
OS MasterMap ontology and a shared vocabulary that contains terms

CORINE LC is a land cover categorization schema intended to provide consistent localized
geographical information on the land cover of the member states of the European Community,
by using satellite data. The CORIINE Land Cover has three hierarchies of categories.
The upper level consists of 5 categories, the middle level of 15, and the lowest one of 44
categories. European Environmental Agency: CORRINE: Land Cover Methodology and
Nomenclature. http://www.eea.europa.eu/publications/COR0-part1, http://www.eea.
europa.eu/publications/COR0-part2

5

GDDD-Geogrpahical Data Description Directory, MEGRIN's GDDD contains information
on the digital geographic information available from Europe's National Mapping Agencies.
(NMAs). Layers names, feature type names and feature attribute types names correspond
to the nomenclature used in the DIGEST Feature and Attribute Coding Catalogue (FACC).
http://www.eurogeographics.org/

6