My friend Ken used to try and get me to open at least one public presentation with a really bad joke: "Metadata. Not an Italian dating service!" Apparently I lack a sense of humor, as I never used it. It does underscore a continuing lack of understanding of what metadata is. Not surprising, really, as we have a distinct lack of term control regarding controlled vocabularies.

The most popular definition of metadata is "data about data." Huh? Think of when you fill out a form - you enter your name, but in two fields: First Name and Last Name. Those field indicators are a kind of metadata, telling the content creator what to enter and helping the content consumer know what to do with the added text string. Have you ever been to a job interview, and been asked to describe yourself in one word? It's pretty hard, and kind of silly too. After all, there are many ways to describe a person - physical characteristics, personality traits, capabilities, and more. Many of those bits of information are incredibly useful in determining if you're a good fit for the job. And so it is with metadata - the extra bits of information that describe a physical or conceptual thing. They can provide context, describe a resource, indicate a process state.

At the 2009 Semantic Technology Conference I had an interesting discussion with Paul Gearon one afternoon, that came to the subject of kinds of metadata. If you were to search for "kinds of metadata," you'll find dozens of definitions. I had it stuck in my head from several years back that there were five, but after thinking about it, I think they can be boiled down to three primary kinds of metadata that are the root containers for all other kinds.

Descriptive

Descriptive metadata describes the "aboutness" of a thing: the subjects it covers, the people or companies it mentions, its creator, title, and more. Library metadata is generally considered to be descriptive.

Structural

Structural or system metadata is more technical in nature: format, mime-type, styling, processing, encryption, authentication, hardware used, software used and storage are examples.

Administrative

Administrative metadata includes information and rules regarding ownership, access, use, copyright, preservation, process and the like.

All kinds of metadata can be used for discovery (search). Application developers should use the kinds that are useful for the task being designed for rather than everything available to reduce noise, processing power and time. Searching for a book, as we know, can be much simpler using Descriptive and Structural metadata such as Author, Title, Subject, Genre, Binding, Condition. Searching for an image is so much more interesting and useful with mime-type (jpg, tiff, png, gif) coordinates, f-stop, lens, series. (See Ian Davis's excellent series on the subject at Synaptica Central, especially this one.) Search from Creative Commons would not be nearly as effective without Administrative Copyright metadata.

Content management systems, web apps and pages, publishing software, eCommerce systems, servers, messaging networks, electronic games, financial networks - everything online uses metadata of some sort. In future posts, we'll talk about many of these, as well as how metadata relates to taxonomies, ontologies, schema, and more.

Additional Reading

Here are some excellent readings on different kinds of metadata:

• NISO. A Framework of Guidance for Building Good Digital Collections Metadata Section. Publication date Thu, 04/17/2008.
• Jeffery, Keith G. Metadata: An Overview and Some Issues Advanced Databases and Metadata; ERCIM News No.35 - October 1998.
• Stanford University Libraries. Assessing Metadata Needs for SUL/AIR Digital Collections: A Guide
• Baca, Murtha. Ed. et al. Introduction to Metadata Online Edition, Version 3.0. Getty Publications. 2008.

Oh, and if you haven't seen this Cat and Girl comic and appreciate sorting stuff by different kinds of metadata, you'll be amused!