Traugott and Dan,
The diagram above shows some first thoughts on an rdf schema for a classification system. The concept node is much the same as for the thesaurus schema, except that there are predicates for what could be called scheme and class or class number.
I haven't bothered to show relationships between concepts from different classification schemes since this should be fairly obvious and would depend on the relationships chosen. The more significant difference from the thesaurus schema involves showing the relationship between the terms connected with the concept and the concept itself. Traugott mentioned the possible need to show the relationships that would be used in an autoclassification system. Here there is a many-to-many relationship between ter m and concept instead of many-to-one as with thesauri. There is also the problem of showing the weighting that is used to indicate the degree of the relationship.
I've simply proposed a new node called 'concept-indicator' to link the term to the concept, this node having properties called weighting, and indicative-term to point to the term, and indicated-concept to point to the concept. A useful extra is that you could add a number of extra properties to the concept-indicator node to show, for instance, what system had been used to determine the relationships and weightings, or maybe also the dataset that had been used for the analysis and date performed.
20th april, 2001