Assisting the Integration of Taxonomic Data: The LITCHI Toolkit

We demonstrate a prototype toolkit that uses constraints and constraint violation repair techniques to enable the automated detection and, where possible, the automated resolution of conflicts in taxonomic databases. 1. The Scientific Problem Taxonomy, the study of variation in organisms, involves the production of taxonomic treatments: these are classification systems in which organisms such as plants and animals are classified. They are hierarchical structures formed from groups known as taxa (singular: taxon). A taxon is known by a scientific name, and refers to a particular group of organisms that exhibit similar characteristics. At the species level a name by which a taxon may be identified is composed of a Latin name and the proposing author, e.g. Vicia lutea Linnaeus. Due to the subjective and cumulative nature of the classification process and geographical or historical differences in background knowledge, many organisms have more than one scientific name. A taxonomist will therefore select one name as the correct name for this taxon, known as the accepted name, while the remaining names are listed as synonyms. A list of the accepted names of the taxa and any accompanying synonyms form a taxonomic checklist. When combining data sets from different sources problems can occur as the treatments may be based on conflicting interpretations and differences of opinion. Treatments are also repeatedly modified to incorporate newly discovered knowledge and to correct problems as they are detected. This can result in a complex history of name changes. Current initiatives aimed at constructing comprehensive biodiversity information systems [2] require the merging or linking of databases. To be able to achieve this merging or linking, any taxonomic conflicts between the databases must be detected and resolved. Currently, resolving conflicts in treatments requires the scrutiny of a taxonomic specialist, and is a time consuming and resource intensive process, particularly as serious difficulties may be encountered when attempting to combine treatments as considerable inconsistencies may exist for some organisms [1,6]. In the LITCHI project (Logic-based Integration of Taxonomic Conflicts in Heterogeneous Information Systems), we have developed a formal model for examining the taxonomic consistency of species checklists, which integrates a number of different kinds of constraint. We have also developed techniques to support repair of violations to these constraints, using knowledge from the domain to guide the effective generation of suitable repair actions. The LITCHI toolkit employs our formal model as a basis for detecting and resolving conflicts in taxonomic checklists.