Metadata: Context Matters

I often think about my former colleague Tom Turner, who died four years ago, in his mid-thirties. In 1997 when I joined Cornell, he was the first true metadata librarian I'd met. From working on various early metadata projects like CUGIR and the USDA Economics and Statistics System, Tom had learned--and taught me--that with metadata, context matters.  Tom's article in the February 2002 issue of Kaleidoscope (the library's staff newsletter) included a number of thoughts I continually return to, like:

Organizing your checking account is not the same as organizing your kitchen cabinet. The same differences apply to metadata; context matters... The use of metadata [is] its fundamental aspect. So we can say that metadata helps you find and/or manage information, serves particular purposes, can be used by people and/or machines, often has structure and/or content rules, and can be created by people or by machines... When beginning a project that involves metadata, the first issue you have to address is what the metadata should accomplish. This will influence what information gets recorded and how it is stored.

So what constitutes quality in metadata can be quite different depending on the communit(ies) being served, the requirements for processing or data exchange, or other factors. Cataloguing is that subset of metadata that serves the communities for which libraries exist.  In addition, cataloguing serves a variety of other audiences internal and external to libraries--and increasingly machines, for system to system exchange, linking, and mashups. 

Last weekend's activities reinforced what Tom taught me about metadata, as I worked with my husband--an independent maker of neon signs and artwork--to set up and "stock" his eBay store. eBay, naturally, runs on metadata; it's the engine that brings buyers and sellers together in a vast, Web-scale network supporting the exchange of goods among ordinary people.  While eBay has set up an impressive framework to house and deploy metadata, the sellers supply the metadata content, and the whole system works quite well.

eBay item listings* and stores have their own sets of metadata elements--all in the context of the metadata's purpose, to bring sellers and buyers together.  In the help pages, eBay emphasizes to its sellers (that is, its metadata creators) the importance of using keywords that will increase the likelihood of an item's appearing in search engine results and 'custom categories' (which support what we would call faceted searching). The categories, a controlled vocabulary set up by the store owner, helps prospective buyers browse the store's virtual shelves.  In less than 24 hours, the individual items in my husband's store were findable in Google, and probably other places I didn't think to look. (Thanks, eBay.)

A seller's purpose is to attract a buyer and close a sale, and that is what eBay metadata is intended to do. This is similar to what library metadata is intended to do--create awareness of an item in a library collection and enable the interested party to get hold of it.  Yet it seems to me that on eBay, the context of metadata use drives what the metadata looks like and how it functions, more than it does in libraries. eBay's contextual deployment of metadata simply works better in the Web 2.0 world that e-commerce sellers and buyers inhabit, where coverage is global; the network is the platform; and metadata is re-mixed, re-used, and syndicated.

In my presentations, I often argue that library metadata practices need to change to put the emphasis on the needs that library metadata meets, rather than the methods used to produce it.  At the present time, too much attention is absorbed by the library metadata framework--that is, the particular methods (AACR2, MARC, the standards for full-minimal-abbreviated records, etc.) we have historically used.  We are paying insufficient attention to the context in which our cataloguing metadata is now deployed, with the result that the libraries' classical metadata framework is losing traction and relevance. 

Of course the library world needs both a rigorous metadata framework and agile responsiveness to the contexts in which our user communities encounter library metdata. It's not my intention to dismiss, minimize, or belittle the importance of a framework for putting library metadata together, but it is a mistake to make the framework an end in itself.  Restoring balance requires a widely-understood, frequently updated understanding of the context--the needs and behaviors and preferences--in which a variety of end user communities use the results of cataloguers' efforts.  What will matter most to the future of libraries is library metadata's suitability in the context of real world use, and not the degree of a cataloguing record's adherence to the specifications of our traditional metadata framework. 

The jury is still out on what "suitability in the context of use" means. There are certainly a lot of opinions and lots of anecdotes to back them up.  Of late, I've been working with a team that is studying the question of what "quality" means for those who use library metadata (including machines). Our team is gaining some surprising  insights into the context of library metadata use on the Web.  Watch for more to come later this summer on our research results.

*eBay Metadata - Item Listings

The metadata elements are (roughly) as follows:
• Title
• Description
• Price (five or more flavors: auction starting price, Buy It Now price, reserve price, shipping costs, tax ...)
• Auction start and ending times
• Shipping details
• Payment method (a controlled vocabulary)
• Item specifics (categories for type of material--another controlled vocabulary)
• Return policy
• A variety of links to related information

This information is remixed with other things, like a digital image of the object for sale.