On Retrieval Guided By Extracted Domain-Specific Knowledge
Troels Andreasen, Rasmus Knappe, Henrik Bulskov · 2005
In this paper we introduce an approach to the modelling of conceptual similarity based on domain knowledge and an approach to aggregation to derive object similarity from concept similarity. Domain knowledge is represented in a special, so-called, domain-specific ontology, which basically is a restriction of a general ontology by a collection of domain concepts or a given document collection. Similarity is derived from the domain-specific ontology and two different variants are considered – an un-weighted and a weighted. Aggregation generalize from concept to object similarity and may be applied in text retrieval to derive answers by comparing query objects with text objects in the base. Adopted for ontology representation is a specific lattice-based concept algebraic language by which ontologies are inherently generative. The modelling of a domain specific ontology is based on a general ontology built upon common knowledge resources such as dictionaries and thesauri. The resulting domain specific ontology and similarity can be applied for surveying the collection through key concepts and conceptual relations and provides a means for topic-based navigation. Keywords: Ontology, Information Retrieval, Fuzzy sets 1