Semantic Web: for the masses, but under the hood.

The article "Semantic Web not for the masses" (Computable dated 27 Feb 2004) makes a remarkably strong statement: "Gartner is very certain that the Semantic Web will never reach a mass audience". This statement is undoubtedly a rather free interpretation of a more balanced conclusion from the Gartner report.

In its current form, this strong statement fits rather well in a long tradition of predictions on IT developments, such as the maximum amount of memory that would be required by anybody: "640K is enough" (Bill Gates in 1981), or the number of computers required worldwide: "about five" (Thomas Whatson, founder of IBM, in 1947).

The name "Semantic Web" alone already conjures up spectacular visions, reinforced by popularising publications [1]. The Semantic Web would, according to [1], enable the computer to understand all our wishes and act upon them. And indeed, this form of the Semantic Web may not be realised in the next decade.

But there are also more realistic definitions of the "Semantic Web". We speak of a "Semantic Web" application when data-exchange happens on the basis of explicit models of the meaning of the data, and when these models are expressed in a standardised language suited for computer processings (such as RDF [2] or OWL [3]).

The introduction of that "Semantic Web" will be gradual, and will certainly not take a decade to happen. And there is a sound economic reason for that: only such semantic technologies will eable the next generation of IT applications. We will give a number of examples, all drawn from current-day Dutch IT industry:

Increases effectiveness of research in pharmacology and reduced costs both require a better use of available knowledge and experience. Elsevier Science owns literally hundreds of thousands of research publications in the pharmacology domain, and sells these to researchers world-wide. Semantic search and browsing of this knowledge [4] enables researchers to localise relevant available knowledge, and increases the use of Elsevier's information.

Techincally perhaps less ambitious, but already opereational as an e-commerce application, is [5]. In this domain, data semantics has been stored for years in relational databases: the relational model is simply a subset of modelling in RDF. At, it would seem that only the user interface is new: the user determines the order of the selection criteria. At each step, the system computes in real time all possible next steps and alternatives. The semantic model ensures that only available holiday combinations are reachable for the user.

Another Semantic Web application close the end user is Internet media. A user would ideally just us a single "tuning knob" to make a choice among the hundreds of thousands of musictracks that are available on-line. Philips already has an internetradio on the market (the FW-i1000). But unfortunately, where in the past a single dial on an old-fashioned radio sufficed to navigate a single linear frequencydomain, the exploration space on the Internet is hyperdimensional and can only be negotiated semantically.

Finally: we too hope that the Semantic Web will never reach the masses! But only in the following sense: very few car drivers are aware of the Nicasil coating in the cylinders of their car, but they are aware of reduced fuel consumption, higher top speeds and extended lifetime of the engine. "Semantic Web" technology is the Nicasil of the next generation of humanfriendly computer applications that are being developed right now.

Frank van Harmelen
Jos van der Meer

Frank van Harmelen is professor of Knowledge Representation at the Vrije Universiteit Amsterdam
Jos van der Meer is founding director of Aduna (

[1] ScientificAmerican: The Semantic Web
A new form of Web content that is meaningful to computers will unleash a revolution of new possibilities, By Tim Berners-Lee, James Hendler and Ora Lassila Scientific American, May 2001



[4] DOPE, Drugs Ontology Project for Elsevier