university of manchester

* Subject Professor Alan Rector from the University of Manchester (UK) presents: "Terminologies and Ontologies: What are they for? What would it mean to QA an ontology (specifically in healthcare?)" * Date Thursday, September 14, 2006 * ONTOLOG Forum session page (with agenda and link to slides) http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2006_09_14 * Abstract (by Alan Rector) Terminologies and 'Ontologies' serve several disparate purposes: o Providing a controlled vocabulary and/or standard set of identifiers o Providing a means of browsing and finding appropriate vocabulary or identifiers o Providing the linguistic terms - synonyms, translations, etc.. - to go with the controlled vocabulary o Cross mapping and translation between different systems o Providing logical criteria which can be used for inference and query expansion o Providing additional 'universal' or intrinsic information about the entities involved o Serving as an index for other background knowledge and resources Most current biomedical ontologies serve primarily the first three functions, with varying efforts towards formal logical criteria. However, there are aspirations, and sometimes claims, for more rigorous functions, and many standardisation efforts pre-suppose more and more formal structure. Each of these functions implies criteria for quality assurance. For example, for managing controlled vocabularies, process issues such as version control and coverage are most critical. Browsing raises issues of human computer interaction, and language raises its own issues. The last three all require a degree of logical coherence and rigour. In addition to many biomedical ontologies, scaling is critical. Biomedical ontologies are large and potentially combinatorially explosive. For some applications, small enumerated terminologies are sufficient. For others, indefinitely large compositional ontologies that cannot, in principle, be pre-enumerated are required. Finally almost all ontologies are based on many poorly articulated assumptions. Any quality assurance methodology must take account both of what can be understood independent of consultation with the originators and what conclusions can be reached after consultation with the originators. Overall we propose an approach to quality along two dimensions - process and content - and a two stage process - the first independent of consultation with the originating authority, the second in consultation with the originating authority.