Friday, June 17, 2005

DCMI & Bibliographic Description

This week the Dublin Core Metadata Initiative published a new recommendation, "Guidelines for Encoding Bibliographic Citation Information in Dublin Core Metadata." I'm finding it to be a muddled mess of possibilities and examples with few real, clear guidelines for an implementer to follow.

The recommendation is described as emerging from the need for describing journal articles in DC. The recommendations tend to center around putting the information that was previously problematic (journal title, volume number, issue number, page range, etc.) within a bibliographicCitation refinement for dc:identifier, while getting the rest of the citation information from other parts of the DC record. "Optionally, but redundantly, these details may be included in the citation as well." This optional part has huge consequences for anyone using DC metadata to get to these citations. One could never know if the complete citation is present in the dc:identifier.bibliographicCitation element, or if one needs to look elsewhere for information to complete the citation. Also, it results in a situation where some of the data needed for this citation is clearly fielded (author in dc:creator, article title in dc:title, etc.), but the rest of it is not. This is hardly an elegant solution to the problem at hand.

Also, "there are no recommendations for providing bibliographic citations in simple Dublin Core." However, it is "suggested" that citation information be put in dc:identifier or dc:description. How is anybody suppposed to use DC for this purpose if the "experts" on it can't bring themselves to turn a "suggestion" into a "recommendation?" This document says to all of us out in metadata-land that there's a solution (actually, TWO solutions - identifier and description - choose between them randomly!), but the powers that be can't or won't formally endorse it, perhaps because it's viewed as a hack. This passive-aggressive "well, we see you have a problem and here are some possible ways to solve it on an official-looking document, but we're not going to tell you that we think any of these solutions are a good idea" crap is really starting to get on my nerves.

I'm also confused about something. bibliographicCitation is a refinement of dc:identifier, and therefore by the DC "dumb-down" rule is a type of identifier. The recommendation says, "dcterms:bibliographicCitation is an element refinement of dc:identifier, recognising that a bibliographic resource is effectively identified by its citation information." But then it goes on to say, "In Dublin Core Abstract Model terms the value of the dcterms:bibliographicCitation property is a value string, encoded according to a KEV ContextObject encoding scheme. It is not intended to be the resource identifier, which for a journal article would probably use an appropriate URI scheme such as DOI." So which is it? Is bibliographicCitation an identifier or not? Is the second quote using "identifier" to mean something different than dc:identifier without telling us? I'm willing to assume for now what I see as a contradiction here comes from my purely surface-level understanding of the DC Abstract model. But maybe not...

No comments: