Abstract: | AbstractNatural history offers an interestingly rich mix of traditional and modern ways of organizing data, information, and knowledge. The Linnaean tradition still defines the basis of how taxonomic knowledge of organisms is organized, while at the same time complementary perspectives on databases and ontologies are developed and implemented, to provide enhanced access to natural history collection data to researchers in taxonomy and biodiversity. Some of this knowledge enrichment may even be automated, and bootstrapped from basic object metadata. This metadata is largely composed of natural language text, which is generally more noisy and ambiguous than numeric data. In this contribution, we present two methods for the automated discovery of metadata from textual object databases: first, the automatic detection of new metadata in existing free-text database columns, and second, the discovery of new ontological relations between metadata elements. |