| Abstract |
Ontology construction is a labour intensive and costly process.
Even though many formal vocabularies are available, creating an ontology
for a specific application is hindered in a number of ways. Firstly,
the process of generating associations is time consuming. Secondly, it
is difficult to keep focus. Third, certain technical modeling constructs
are hard to understand. We propose ROC as a method to cope with
these problems. ROC builds on the well-known Methontology approach.
We reuse existing sources to create a so-called proto-ontology. Rather
than using multiple intermediate representations to interact with the
domain expert, we present ‘natural-language-like’ statements generated
from RDF-based triples. Moreover, we strictly separate the roles of problem
owner, domain expert and knowledge engineer. They keep focus by
monitoring a well-defined application purpose. ROC assists in efficiently
putting forward all relevant concepts and relations by providing a large
but task-specific set of potential associations. We have implemented an
initial set of tools to support ROC. This paper describes the ROC method
and two application cases in which we evaluate the overall approach.
|