Improving Domain Ontologies by Mining Semantics from Text

Dittenbach, M., Berger, H. and Merkl, D.

    The creation and maintenance of domain ontologies is a costly and time-consuming task. With the advent of ontologies being used in many different fields of computer science, developing appropriate algorithms and methods to support or automatize ontology engineering have become an increasingly important goal. Hence, we present a connectionist approach to visualize semantic relations inherent in free-form text documents related to a specific domain. In particular, we exploit word co-occurrences to capture relatedness of words in order to generate numeric representations of the words' contexts. We use the self-organizing map, a well-known neural network model with unsupervised learning function, to map the high-dimensional data onto a two-dimensional representation for convenient browsing. This intuitive view on the domain vocabulary supports the construction and enrichment of domain ontologies by making relevant concepts and their relations evident. We underline this approach with an example from the tourism domain.
Cite as: Dittenbach, M., Berger, H. and Merkl, D. (2004). Improving Domain Ontologies by Mining Semantics from Text. In Proc. First Asia-Pacific Conference on Conceptual Modelling (APCCM2004), Dunedin, New Zealand. CRPIT, 31. Hartmann, S. and Roddick, J. F., Eds. ACS. 91-100.
pdf (from crpit.com) pdf (local if available) BibTeX EndNote GS