Databases Reference
In-Depth Information
In this chapter, we explore research and tools that support the visual and inter-
active ontology matching process. We begin by discussing the cognitive difficulties
with creating an ontology matching (Sect.
2
). In Sects.
3
-
5
, we discuss interactive
tools for ontology matching, schema matching, and Web 2.0 approaches. In Sect.
6
,
we present several user-oriented evaluations and experiments that researchers in this
area have carried out. We discuss common themes in Sect.
7
, challenges and future
directions for this field in Sect.
8
. We conclude the chapter in Sect.
9
.
2
Why is Ontology Matching Difficult?
Reconciling different ontologies and finding correspondences between their con-
cepts is likely to be a problem for the foreseeable future. In fact, every self-
assessment of database research has listed interoperability of heterogeneous data
as one of the main research problems [
Bernstein and Melnik 2007
]. Despite years
of research on this topic, ontology and schema matching is far from being a fully
automated task. In general, a user must interact with an ontology-matching tool
to examine
candidate matchings
produced by the tool and to indicate which ones
are correct, which ones are not, and to create additional correspondences that the
tool has missed. However, this validation process is a difficult cognitive task. It
requires tremendous patience and an expert understanding of the ontology domain,
terminology, and semantics.
Obtaining this understanding is very difficult. Languages are known to be
locally
ambiguous
, meaning that a sentence may contain an ambiguous portion that is no
longer ambiguous once the whole sentence is considered [
PPP 2006
]. Humans use
detailed knowledge about the world to infer unspoken meaning [
NLP 2002
]. How-
ever, an ontology often lacks sufficient information to infer the intended meaning.
The concepts are largely characterized by a term or a small set of terms, which may
be ambiguous.
The underlying data format that is used for specifying the ontology also intro-
duces potential problems. The language used (e.g., OWL, RDF, XSD) constrains
the expressiveness of the data representation. For example, many formats lack
information relating to units of measure or intended usage [
Bernstein and Melnik
2007
].
Ontologies are also developed for different purposes and by users with potentially
opposing world views or different requirements. As a result, two ontologies may
describe the same concept with different levels of granularity or the same concept
with different intended application or meaning. All of these issues make discovering
and defining matchings a very challenging problem for both man and machine.
As a consequence, to create accurate matchings in a reasonable amount of time,
users and tools must be paired together. This process, usually referred to as
semi-
automatic ontology matching
, typically follows an iterative process that is similar
to the one that we describe in Fig.
2.1
. Recently, this approach has received greater
attention and an increasingly larger number of semiautomatic tools are becoming
