Mappings between TEI Header elements and Dublin Core fields have for the most part been stable and well identified for many years. There are however a few cases where we could helpfully provide more guidance, especially as the DCMI concepts are now clearly defined and standardised and even (I believe) mapped to eg DCR concepts.
This is a particularly striking example. We suggest for teiHeader/@type only two possible values "text" and "corpus". It would be more helpful to explicitly refer to http://dublincore.org/documents/dcmi-type-vocabulary/ which proposes and defines "Collection", "Dataset", "Event", etc. We should either simply reference this list from the spec or reproduce it as a suggested vallist.
Assigning to Paul Schaffner to triage and report to council.
Council at F2F Nov 14 decided that rather than add values to @type on teiHeader, the attribute was no longer needed at all and should be deprecated. Deprecated at revision 13064, and the six examples in the guidelines removed. Files changed: Guidelines/en/CC-LanguageCorpora.xml, Guidelines/en/HD-Header.xml, and Specs/teiHeader.xml.
Last edit: Paul Schaffner 2014-11-19
Most intriguing, this.
@type on teiHeader helps a lot in any non-primitive system of headers for corpora, whereby we can distinguish between, for example, corpus-level headers and text-level headers.
See e.g. http://nlp.ipipan.waw.pl/TEI4NKJP/
Free-standing headers might benefit from being distinguished machine-readably from "regular" headers of any sort.