For this project we propose the use of a subset of the SKOS rules modified to deal with the semantic heterogeneity stemming form differences in language terminology domain coverage granularity. We provided some examples and guidelines on how to apply the rules. In addition we also introduced the notion of inheritance.
This work is just at the beginning. At this time, we can concentrate more on the design of the mapping procedure and the testing phases. As we perform the work we are likely to encounter other unanticipated problems and issues that will require revision of our initial guidelines. We also believe that this work can benefit other terminology mapping projects especially by contributing to the discussion concerning the refinement of the SKOS specification based on its application to real world data.
As a mapping procedure involves a great number of terms and mapping relations, it is nearly impossible to do the job without the help of computers. For instance automatic matching of CAT and AGROVOC terms with the same English and Chinese can be proposed for exactMatch mappings. For concepts that are not mapped an automatic procedure can be applied to check if an ancestor has been mapped and if not the concept can be flagged for manual mapping. The use of tools such as Protégé and Prompt can also provide semi-automatic assistance on the mapping.
