6
Automated Geographical Information Fusion
and Ontology Alignment
Matt Duckham
1 and Mike Worboys
2
1 University of Melbourne, Melbourne (Australia)
2 University of Maine, Orono (USA)
6.1 Introduction
Geographical information fusion is the process of integrating geographical information from diverse sources to produce new information with added value, reliability, or
usefulness (cf. [14, 67]). Geographical information fusion is an important function of
interoperable and Web-based GIS. Increased reliance on distributed Web-based access to geographical information is correspondingly increasing the need to efficiently
and rapidly fuse geographical information from multiple sources.
The overriding problem facing any geographical information fusion system is
semantic heterogeneity, where the concepts and categories used in different
geographical information sources have incompatible meanings. Most of today’s geographical information fusion techniques are fundamentally dependent on human
domain expertise. This chapter examines the foundations of automated geographical
information fusion using inductive inference. Inductive inference concerns reasoning from specific cases to general rules. In the context of geographical information
fusion, inductive inference can be used to infer semantic relationships between categories of geographical entities (general rules) from the spatial relationships between
sets of specific entities. However, inductive inference is inherently unreliable, especially in the presence of uncertainty. Consequently, managing reliability is a key
hurdle facing any automated fusion system based on inductive inference, especially
in the domain of geographical information where uncertainty is endemic.
This chapter develops a model of automated geographical information fusion
based on inductive inference. Central to this model are techniques by which unreliable inferences and data can be accommodated. The key contributions of this chapter
are to
• define the foundations of automated geographical information fusion using inductive inference;
• explore some of the limitations of automated geographical information fusion,
inherent in inductive inference;
Automated Geographical Information Fusion
and Ontology Alignment
Matt Duckham
1 and Mike Worboys
2
1 University of Melbourne, Melbourne (Australia)
2 University of Maine, Orono (USA)
6.1 Introduction
Geographical information fusion is the process of integrating geographical information from diverse sources to produce new information with added value, reliability, or
usefulness (cf. [14, 67]). Geographical information fusion is an important function of
interoperable and Web-based GIS. Increased reliance on distributed Web-based access to geographical information is correspondingly increasing the need to efficiently
and rapidly fuse geographical information from multiple sources.
The overriding problem facing any geographical information fusion system is
semantic heterogeneity, where the concepts and categories used in different
geographical information sources have incompatible meanings. Most of today’s geographical information fusion techniques are fundamentally dependent on human
domain expertise. This chapter examines the foundations of automated geographical
information fusion using inductive inference. Inductive inference concerns reasoning from specific cases to general rules. In the context of geographical information
fusion, inductive inference can be used to infer semantic relationships between categories of geographical entities (general rules) from the spatial relationships between
sets of specific entities. However, inductive inference is inherently unreliable, especially in the presence of uncertainty. Consequently, managing reliability is a key
hurdle facing any automated fusion system based on inductive inference, especially
in the domain of geographical information where uncertainty is endemic.
This chapter develops a model of automated geographical information fusion
based on inductive inference. Central to this model are techniques by which unreliable inferences and data can be accommodated. The key contributions of this chapter
are to
• define the foundations of automated geographical information fusion using inductive inference;
• explore some of the limitations of automated geographical information fusion,
inherent in inductive inference;