One RDF approach to semantic drift is to situate a vocabulary among other terms.
TMDM topic maps enable a user to gather up information that they considered as identifying the subject in question.
Additional information helps to identify a particular subject. (RDF/TMDM approaches)
Isn’t that the opposite of semantic drift?
What’s happening in both cases?
The RDF approach is guessing that it has the sense of the word as used by the author (if the right word at all).
Kelb reports approximately 48% precision.
So in 1 out of 2 emergency room situations we get the right term? (Not to knock Kelb’s work. It is an important approach that needs further development.)
Topic maps are guessing as well.
We don’t know what information in a subject identifier identifies a subject. Some of it? All of it? Under what circumstances?
Question: What information identifies a subject, at least to its author?
Answer: Ask the Author.
Asking authors what information identifies their subject(s) seems like an overlooked approach.
Domain specific vocabularies with additional information about subjects that indicates the information that identifies a subject versus merely supplemental information would be a good start.
That avoids inline syntax difficulties and enables authors to easily and quickly associate subject identification information with their documents.
Both RDF and TMDM Topic Maps could use the same vocabularies to improve their handling of associated document content.