Metadata: A fundamental component of the Semantic Web

Bulletin of the American Society for Information Science & Technology, Apr/May 2003 by Greenberg, Jane, Sutton, Stuart, Campbell, D Grant

In their widely discussed May 2001 article on the Semantic Web in Scientific American (www. sciam.com/article.cfm?articleID=00048144-10D21C70-84A9809EC588EF21) Tim Berners-Lee, James Hendler and Ora Lassila present a scenario in which a person named Pete is listening to the Beatles through his home entertainment center. Lucy, Pete's sister, phones from the doctor's office to explain that their mother needs a series of biweekly physical therapy sessions. The first few paragraphs of this article tell how both Pete's and Lucy's Semantic Web agents (hereafter referred to as agents) communicate with each other and transverse the Semantic Web to schedule their mother's physical therapy session, how Pete is not pleased with the initial plan and how later that evening Pete sends his agent back onto the Semantic Web to find an alternative plan. Pete's Web agent completes this second task and reschedules several of his personal and less important work appointments.

Realizing this scenario is dependent not only on the ability of Pete's and Lucy's agents to communicate with each other, but their ability to transverse a system of structured semantic knowledge that is forming the Semantic Web. This system of semantics is metadata. With efforts to build the Semantic Web, we are beginning to see the metadata infrastructure that agents need to carry out tasks that the initial Web has not been able to support. To this end, implementing and harvesting metadata is fundamental to the success of the Semantic Web. This article provides an overview of metadata as a key component of the Semantic Web - its vision and architecture; metadata vocabularies; enabling technologies; and authoring and annotation.

Semantic Web: Vision & Architecture

Clearly articulated visions and architectural plans, drawn by great thinkers and experts of the time, form the underpinnings of many of the world's most significant structures. Consider the Panama Canal. Crossing the mass of land called the Americas at its narrowest point by means of a waterway was a vision shared by many throughout history, including indigenous people of the Americas, Christopher Columbus and merchants worldwide. Execution of architectural plans, first by leading French engineers in 1879 and then by a U.S. Commission, led to completion of the Panama Canal in 1914. In like fashion, an evolving and shared vision, supported by an architectural plan, underlies the development of the Semantic Web.

The Vision. The Semantic Web was envisioned by Tim Berners-Lee, inventor of the World Wide Web (Web), and is now being further defined by researchers and visionaries, but it was inspired by a host of creative thinkers who have, throughout history, looked to technological innovation as a way not only to control, but also to transform the world's mass of information into intelligence. Vannevar Bush was one early pioneer in this area with his vision of the Memex (www.the atlantic.com/unbound/flashbks/computer/bushf.htm) - a mechanism using associative indexing to link the world's vast body of scientific recorded knowledge to discover new knowledge. Another significant idea is Alan Turing's conceptualization of the Turing Machine and its use of logic to transform numbers into intelligence. (See www.turing.org.uk/ turing/scrapbook/machine.html.) The vision supporting the Semantic Web draws upon these ideas and new ideas inspired by technological developments to create intelligence.

The Architecture. The Semantic Web's architecture, captured by Berners-Lee, is represented in Figure 1. Each layer supports or has a connection to metadata.

* URIs and Unicode. URIs (uniform resource identifiers) are unique identifiers for resources of all types - from schemas to people. A major component of the base layer, URIs are metadata and function like ISBNs (international standard book numbers) or Social Security numbers in the context of the Web.

XML NS XMLschema. Extensible Markup Language (XML) and more recently XML schemas facilitate the creation, use, and syntactic interoperability of metadata vocabularies. NS (namespaces), which are identified via URIs, secure semantic interoperability among metadata vocabularies.

RDF and RDFschema. The RDF family further supports interoperability at the semantic level. RDF developments comprise the base Web language, so that agents, like Pete's and Lucy's discussed above, can make logical inferences, based on metadata, to perform tasks.

Ontology vocabulary. Ontologies are metadata systems (referred to as metadata vocabularies in this article). The ontology layer represents the Semantic Web's central metadata artery, where simple descriptive to complex classificatory schemas are to be created and registered so that agents can intelligently interpret data, make inferences, and perform tasks. Jacob's article in this issue discusses ontologies in detail.

Logic. We make logical inferences in our performance of daily tasks. For example: If N denotes new unread email in an email inbox, then if an N appears by a particular message, the message is new unread email. This inference is based on evidence provided by the letter N. The logic layer of the Semantic Web works on this basic principle through First Order Predicate Logic. An agent can derive a logical conclusion (or reason) in the process of completing a task based on what are essentially "facts" rendered from semantically encoded metadata. Other types of logic may also be applicable in the Semantic Web.