Semantically processing parallel colour descriptions

Information integration and retrieval are useful tasks in many information systems. In these systems, it is far from an easy task to directly integrate information from natural language (NL) sources, because precisely capturing NL semantics is not a trivial issue in the first place. In this paper, we choose the botanical domain to investigate this issue. While most existing systems in this domain support only keyword-based search, this paper introduces an ontology-based approach to process parallel colour descriptions from botanical documents. Based on a semantic model, it takes advantage of ontologies so as to represent the semantics of colour descriptions precisely, to integrate parallel descriptions according to their semantic distances, and to answer colour-related species identification queries. To evaluate this approach, we implement a colour reasoner based on the FaCT-DG Description Logic reasoner, and present some results of our experiments on integrating parallel descriptions and species identification queries. From this highly specialised domain, we learn a set of more general methodological rules.