> Quoting Jonathan Rochkind <[email protected]>: > >> I'm confused why you need an entity for 'the whole thing'. I suggest >> that any assertions you think you want to make on 'the whole thing' are >> better made on a particular Work, and I suggest that's the intent of the >> FRBR model. The Work entity already is best thought of a set including >> all of it's EMI (a way of thinking not in the FRBR document, but >> _entirely_ consistent with it) -- any assertions on the Work already are >> on 'the whole thing'.
I copied this statement to Gordon Dunsire, who is doing the RDF representations of the FR's for IFLA. His response is that FRBR is a "bottom-up" model: Quoting Gordon: "In the OWL ontology I have for FRBR, the WEMI dependencies are one-way (bottom-up); there is no constraint on W that says there must be an E (but there is a constraint on E saying there must be one and only one W). Of course all assertions are ultimately on the Work: any assertion on an E is also an assertion by implication on W. But it is insufficient to model this with just W as the domain of all assertions." It seems that having the URI for a Work therefore requires something approximating a query of the type: retrieve all expressions that express this work, then retrieve all manifestations that manifest those expressions, and all items that physically embody that manifestation. To me this parallels the act of cataloging (with the item in hand), but is the reverse of the user's retrieval path. The bottom-up view should be easier because the links between I&M, M&E, E&W must be explicit in the data. Gordon also says: "The separate elements of WEMI are in a hybrid mono/poly hierarchical relationship, and not just an equal-level designation of sets of statements/attributes/assertions. In other words, WEMI is take-it-or-leave-it." My dilemma is therefore that any statement I wish to make has to be made on some identified entity, that is on a "thing" with a URI. So if, for example, I want to say: "this X is more popular than that Y" then I have to choose a FRBR entity for X and for Y; I have to know if X and Y are Works, Expressions or Manifestations. To me, this is a fairly high hurdle for data sharing because I don't expect most bibliographic data to be in a frbr-ized format. If I have something like a MARC record that is a single bibliographic unit, and I want to make a connection between a MARC "thing" and a FRBR "thing", to begin with I cannot use any of the FRBR-defined predicates because the FRBR predicates are all specific to one of the WEMI entities (although it isn't clear if the subject and object must be of the same FRBR entity), and there is no URI that represents the whole WEMI. This means that: 1) none of the FRBR predicates can be used to express a relationships between a FRBR entity and a non-FRBR bibliographic entity 2) if one should invent a predicate that can be applied to whole bibliographic entities, any non-FRBR bibliographic entity that is a subject can only have as its object either another non-FRBR bibliographic entity, or one specific FRBR entity (W or E or M or I). Therefore in such a statement the subject and object of the triple will be quite different in nature, one being a general bibliographic entity, and the other being a specific FRBR entity. The full bibliographic description of the FRBR-ized "thing" can be recreated, but only specific WEMI can be the subject or object of a description. What I guess I would like would be a way to identify the full Group 1 so that they can be included in statements relating to non-FRBR-ized bibliographic data. I admit that there is also a question of the other 2 FRBR groups and how they would be included in this picture. The analogy of Group 2 to name authority and Group 3 to subject authority is not quite correct; authority data exists separately from bibliographic description in a system where the actual authoritative headings are included in the bibliographic description. That is not true for FRBR: the Work entity, as defined, does NOT have properties for creators and subjects. Creators/agents and subjects are separate entities that have a relationship with the Group 1 entities but aren't included in them. All 3 FRBR groups are necessary to create a complete bibliographic description (presuming that creators/agents and subjects exist for that description). Sorry if this has gotten complex, but I think it IS complex and needs to be sorted out. Meanwhile, it looks like the RDF output from OL will need to take the form: WorkURI work properties ExpressionURI (=ManifestationURI + "#Expression") language format expressionof:WorkURI ManifestationURI manifestation properties manifestationof:ExpressionURI ExpressionURI (=ManifestationURI + "#Expression") language format expressionof:WorkURI ManifestationURI manifestation properties manifestationof:ExpressionURI etc. I'm not sure of the use of the hash segment on the URI, but have tested it an it returns the OL book record. It does not limit to the Expression data, which is mixed in with the Manifestation data in the OL book record. I would like advice on whether this is an appropriate use of the # in an RDF URI. Thanks, kc -- Karen Coyle [email protected] http://kcoyle.net ph: 1-510-540-7596 m: 1-510-435-8234 skype: kcoylenet _______________________________________________ Ol-tech mailing list [email protected] http://mail.archive.org/cgi-bin/mailman/listinfo/ol-tech To unsubscribe from this mailing list, send email to [email protected]
