- From: Leonard Will <L.Will@willpowerinfo.co.uk>
- Date: Sun, 8 Jun 2008 16:34:07 +0100
- To: Antoine Isaac <Antoine.Isaac@KB.nl>
- Cc: Simon Spero <ses@unc.edu>, public-esw-thes@w3.org
On Sun, 8 Jun 2008 at 15:53:47, Antoine Isaac <Antoine.Isaac@KB.nl> wrote >My problem is that if you add the axiom Alistair proposes, then a >knowledge base that has both SKOS information and dc:subject would >return, in a set of indistinguishable results, both the dc:subject 'at >the right level' and the more general subjects that can be inferred >from it. >Or more precisely: you can make the distinction, but then your query to >the knowledge base should be different, more explicit, like 'find all >the subjects of the book for which there is no specialization which is >also a subject of a book'. > >It could be perfectly valid from your point of view, but I rather see a >search system activate itself some useful semantics (by explicitly >asking for all the books indexed by a subject and all its descendants, >which would give animals and platypus and so on) rather than having it >eventually to "de-activate" some semantics which were assumed as a >default, and therefore did not apply to all cases (e.g. for a query >like "I just want to display/access the subjects of this book as they >were defined by a professional indexer") I think the search system should give the user the option of retrieving only documents indexed with the exact term asked for, or that term and its descendants, or of broadening the search by looking at ancestor terms too. This is the sort of interaction that Doug Tudhope's group in Glamorgan are developing. The default behaviour should be able to be set by the user, so it is not a case of "deactivating semantics" but rather of switching the system as required. I suppose that the "semantic web" aim is to take the user out of the loop and let the system make such decisions, but I strongly believe that interaction with the user will give better results at the present level of development. Though there is an attitude that users can only be expected to put one or two terms into a Google-type box, I think there is a place for a system that shows them options and lets them see relevant bits of the subject hierarchy from which they can choose search terms. This needs excellent interface design, though, to be helpful and not off-putting. That is outside the scope of the SKOS discussion, I imagine. >> > The indexing term "animals" would be assigned to documents which either >> > >> > a. Deal with animals in general >> > b. Deal with several types of animal, too many to index individually >> > c. Deal with a type of animal for which no more specific term exists in the >> > thesaurus. >> >> >> The only requirement placed on the syndetic structure is support for >> hierarchical relationships between terms such that the narrower term is a >> proper subset of the broader one; otherwise upward posting and the rule of >> three (for suitable values of three) don't work, and the use of specific >> terms will cause recall to drop. Unfortunately, SKOS as it now stands >> doesn't support this kind of relationship. > >Well maybe I don't understand your comment, (The "rule of three" is library jargon for the convention that if a document deals with several (e.g. more than three) sibling concepts it should be indexed by the parent concept which they share, i.e. my case b. above. ) >but it seems to me that skos:broaderTransitive would allow for getting >results without the recall drop. I'm not quite sure how you are suggesting that this should be used, but I take it that it means that a thesaurus hierarchy should not mix types of relationship such as genus/species and whole/part, so that the BT/NT relationships are transitive throughout. Good advice, but as Simon says it is not enforced in SKOS because of the need to accommodate existing less well-structured KOSs. Leonard [I am sending this only to public-esw-thes@w3.org and not copying it to your personal addresses. I have been receiving four copies of most messages recently: one personally addressed and one from the list, each one having a redundant HTML part as well as the plain text. :-) ] -- Willpower Information (Partners: Dr Leonard D Will, Sheena E Will) Information Management Consultants Tel: +44 (0)20 8372 0092 27 Calshot Way, Enfield, Middlesex EN2 7BQ, UK. Fax: +44 (0)870 051 7276 L.Will@Willpowerinfo.co.uk Sheena.Will@Willpowerinfo.co.uk ---------------- <URL:http://www.willpowerinfo.co.uk/> -----------------
Received on Sunday, 8 June 2008 15:36:45 UTC