The 13th European Networked Knowledge Organization Systems (NKOS) Workshop that took place in London last week had a special session on mapping between vocabularies, so it was good to be able to present our current thoughts on this topic. Amongst the 20 or so delegates were Stella Dextre Clarke, who chaired the working group that developed the ISO standard, and Andreas Kempf, our friend and colleague from GESIS.
As a result of our recent thesaurus alignment work we defined the relationship between core concepts in HASSET and ELSST as a type of mapping. Core concepts must share the same preferred term labels and BTs, but may vary in all other metadata. Where scope notes and sources are also identical, core concepts are said to be ‘exact’ equivalents, otherwise, if scope notes or sources differ, they are said to be ‘close’ equivalents. These are local definitions which are used primarily for term management purposes. In the interests of interoperability, however, it is useful to be able to relate these definitions to the definition of inter-vocabulary equivalence in the ISO 25964-2 guidelines and in SKOS.
The semantics of equivalence mappings in ISO and SKOS was raised at the Workshop in a number of presentations. For example, Ceri Binding from the University of South Wales discussed mapping work that he and his colleagues have undertaken within the SENESCHAL project between cultural heritage thesauri – The Royal Commission on the Ancient and Historical Monuments of Scotland (RCAHMS )and the Royal Commission on Ancient & Historical Monuments of Wales (RCAHMW ). Like HASSET and ELSST, these thesauri are very similar and have a common ancestor (the English Heritage Monument Types Thesaurus), and may differ only slightly in their scope notes.
Johan De Smedt of TenForce meanwhile presented work being done within the EU-funded European Skills, Competences, Qualifications and Occupations (ESCO) initiative, which aims to create a hub vocabulary on these topics for the EU labour market. Again, the meaning of equivalence was discussed, this time in a multilingual context, since many occupations require qualifications that have no direct equivalents across countries.
Putting it all in context
It was agreed that the definition of equivalence in ISO has no clear criteria, and is context- dependent. In other words, a mapping can be declared to be an exact equivalence if it would serve users’ interests to do so. What is important for information retrieval is not the similarity of the concepts per se, but the similarity of the results found by applying them to a given task. However, as Stella pointed out, the fact that the definitions of mappings are context-dependent causes a problem when the mappings are released to the wider world as linked data, since the purpose to which they will be put is not known. Everyone agreed that it would be useful to provide mappings with metadata to explain how and why they were established.
We will take a closer look at scope notes in HASSET and ELSST in our future development work. In the meantime, our presentation can be viewed, together with other workshop presentations on the NKOS website and on the UK Data Service website.