Con ict Detection for Integration of Taxonomic Data

Over recent years, international initiatives such as the 1993 U.N. Convention on Biological Diversity have highlighted the need for information about species diversity on a global scale. However, attempts to build global information systems by integrating smaller, independently created biodiversity databases have been hampered by diierences in the sets of species names used. Some databases use diierent names to refer to the same species, while in other cases the same name can be applied to diiering deenitions of a species, or even entirely diierent species. The LITCHI project aims to assist biologists in the integration of databases by searching for connicts within taxonomic checklists (i.e. lists of the species names used in a database and the relationships between them). In order to detect such connicts, we have created a formal model of taxonomic practice, which describes (amongst other things) what it means for a checklist to be consistent and well-speciied. This model has been used as the basis for a prototype tool that uses Prolog to search for naming connicts within a relational database of checklists. In this paper, we describe the background to our formal model and show how it has been used to implement the LITCHI system. Our prototype tool is already proving its worth by detecting con-icts and errors within real taxonomic checklists.