Microsoft Word 16 ok6408. doc

Key words: ontology, catalog, database interoperability. 
1. INTRODUCTION 
The growing number of independent object sources, easily accessible 
through the Web, has only exacerbated the heterogeneity problem. More 
precisely, accessing independent object sources requires remapping 
structurally heterogeneous objects, that is, objects that are organized 
following different conceptual schemas, as well as semantically 
heterogeneous objects that is, objects with distinct meanings. Structural and 
semantic heterogeneity problems are challenges that the distributed 
databases community has long faced (Özsu and Valduriez, 1999). 
To enable interoperability, remote systems must be able to locate and 
access object sources, and to interpret and process the objects. One solution 

---

236 
proposed by the database community relies on generating mappings between 
pairs of conceptual schemas. However, this pairwise mapping between 
schemas becomes impracticable if the number of object sources is fairly 
large. Another solution consists in adopting a global schema. In this case, 
each conceptual schema has to be mapped into the global schema. Yet 
another approach, advocated more recently, is to use ontologies to expose 
implicit knowledge (Wache et al., 2001; Uschold and Grüninger, 2001) 
thereby enabling interoperability. 
Mena et al. (2000), proposes a semantic based integration approach that 
uses multiple ontologies, instead of an integrated view. In this context, 
ontologies are virtually linked by interontology relationships, which are then 
used to indirectly support query processing.
Reed and Strongin (2004) propose a new service for generalized 
distributed data sharing and mediation using XRIs (eXtensible Resource 
Identifiers). The goal of XDI is to enable data from any data source to be 
identified, exchanged, linked and synchronized into a machine-readable 
dataweb using XML documents. 
However, independently of the approach adopted to map schemas or 
ontologies, it might be impossible to define mappings between objects from 
distinct sources. For instance, consider two object sources about enterprise 
installations, maintaining information about buildings, industrial 
installations, etc. Suppose that each source uses its own installation 
identifier, say, one uses the address as identifier, and the other uses an 
installation code. Suppose also that neither stores both the address and the 
code of the installations. It is then obvious that, given an installation 
identified by its address, it becomes impossible to locate the same 
installation by its code, and vice-versa. Note that this is true even if one 
aligns the objects classes in both sources. In fact, without explicit mappings 
between the object instances, these two data sources cannot interoperate. 
This problem can only be addressed if an object catalog is defined that 
explicitly stores object instance mappings.
This paper then proposes the concept of Ontology-based Geo-Object 
Catalog (OGOC), as a strategy to address the interoperability problem 
between geographic object sources. We mean by geographic object, or geo-
object, any data that has some information about its spatial location (we will 
avoid using the term “feature” in this paper). The OGOC will act as a 
generalized mediator for a federation of geo-object sources, providing 
services to access and search for federated data and metadata. To meet this 
requirement, the catalog will store: (1) a reference ontology, similar to a 
global conceptual schema; (2) local ontologies describing object sources; (3) 
ontology mappings from the local ontologies to the reference ontology; (4) 
sets of instances of geo-objects, acting as standard geo-objects: (5) instance 

---

237 
mappings from reference geo-objects to the geo-objects stored in each 
source. 
In short, an OGOC enables interoperability among the federated geo-
object sources on both the data and the metadata levels. 
The OGOC concept generalizes and combines the OpenGIS Consortium 
(OGC)
1
catalog and gazetteer notions, referenced in this text, respectively, as 
OGC Catalog and OGC Gazetteer.
An OGC Catalog is a collection of descriptive information (metadata) 
about data stored in a geographic database (Nebert, 2002). Thus, metadata 
describes the properties that can be queried and requested through catalog 
services. An OGC Catalog provides discovery, access and management 
services, allowing the user to locate and modify metadata, and to request 
services on the data. 
The OGC Gazetteer (Atkinson and Fitzke, 2002) is a spatial dictionary of 
objects with geographic attributes. Each instance of a gazetteer service 
typically covers a limited region, such as a country, and has an associated 
vocabulary of geo-object identifiers. An OGC Gazetteer provides operations 
to retrieve: 
• the service description and geo-object types that it can handle 
(getCapabilities); 
• the schema definition of a geo-object type (describeFeatureType); 
• sets of geo-objects (getFeature). 
This paper is organized as follows. Section 2 introduces the concept of 
Ontology-based Geo-Object Catalog. Section 3 presents a generic 
architecture for an OGOC, briefly introduces a framework to generate 
customized OGOCs, and describes scenarios where an OGOC can profitably 
be used. Section 4 addresses additional functionalities that an OGOC should 
have. Finally, Section 5 contains the conclusions and suggestions for future 
work.