Workshop on The Digitization of Language Data: The Need for Standards
21-24 June 2001, Santa Barbara, CA

Metadata Working Group: Experience Report

Gary Simons, SIL International
June 14, 2001

Each member of the Metadata Working Group has been asked to apply the OLAC or ISLE metadata to some of their own material and comment on how well it worked. I have implemented a mapping of the catalog database for SIL International's Language and Culture Archives onto OLAC metadata descriptions. At this point, our archives consist of approximately 30,000 publications in or about 1832 languages of the world. Here is an example of an OLAC metadata description for item 22059 in the archives:

<olac xmlns="http://www.language-archives.org/OLAC/0.3/" >
  <creator>Derbyshire, Desmond C.</creator>
  <date code="1986"/>
  <title>Topic continuity and OVS order in Hixkaryana</title>
  <source>In Joel Sherzer and Greg Urban (eds.), Native South American
     discourse , 237-306. Berlin: Mouton.</source>
  <type code="Text" />
  <type.data code="description/grammatical" />
  <subject>Word order</subject>
  <subject>Topic</subject>
  <subject>Typology</subject>
  <subject.language code="x-sil-HIX"/>
  <identifier>http://www.ethnologue.com/show_work.asp?id=22059</identifier>
</olac>

The existing data generally map well into the elements of the OLAC metadata set, but there are some problem areas:

The definition of the <source> element in the OLACMS document needs to be refined. It, like the Dublin Core definition it follows, is still too vague. In this example, <source> seems like the appropriate element for the citation of the larger work in which the article is found, but that is not clear from the definition.
I added the <type.data> classification by hand. A significant portion of these could be assigned automatically from information in the catalog database, but to handle every case, we would need to add information to our database.
The identifier is for an online description of the item on the soon-to-be-unveiled www.ethnologue.com web site, rather than for the content of the item. This may be a significant enough distinction that we would want to add a coded refinement to make this distinction.

I have also implemented the generation of DC and OLAC metadata description for each language entry in the new online edition of the Ethnologue, so these can be accessed via the Open Archives Initiative protocol. For example,

<olac xmlns="http://www.language-archives.org/OLAC/0.3/" >
  <date code="2000"/>
  <title>GHOTUO: a language of Nigeria</title>
  <description>A page from the Web edition of the Ethnologue (14th edition)
     giving basic facts about the language and where it is spoken.</description>
  <publisher>SIL International (www.sil.org)</publisher>
  <subject.language code="x-sil-AAA"/>
  <type code="Text" />
  <type.data code="description" />
  <format code="text/html"/>
  <identifier>http://www.ethnologue.com/show_language.asp?code=AAA</identifier>
</olac>