You weren't properly introduced to Semantic Web. It is not an approach to solve a problem, but an idea which makes lot of problems possible or simpler (and some more difficult) to solve.

If I could introduce you the Semantic Web in one sentence, I would say it is a group of technologies, mainly developed by W3C, aimed for representation of information in a form understandable by machines on global web scale.

Anyway James Hendler has a really nice informative post about Semantic Web and Linked Data: What is the Semantic Web really all about?

It is always interesting to observe that an abstraction is frequently equated to over-engineering.

Take the case of XHTML (why we go now to HTML5 is not related to this discussion), by using the XML syntax it made things simpler: it just relaxed the parsing requirements by adhering to simpler rules, with several benefits, from lower memory and CPU usage.

But it did not made things easy for authors, and it felt over-engineered especially because why it was made so wasn't communicated clearly.

You can say that RDF relaxed the syntactic requirements in exchanging data, through triples and URIs.

While "the other things on top" of RDF: RDFS, OWL, etc. (see the semantic web cake) are filling up towards the Semantic Web idea.

Now, the fact the the Semantic Web is perceived as over-engineered is another communication issue. And this is not simple issue, as we don't quite know where are going. We discover daily—as we publish and link new data—unforeseen possibilities.

All that we have are a set of strong principles—grounded in the proven Architecture of the WWW—which we try to follow almost religiously and this fact, those principles/axioms which do not look practical (considering the current problems, that can be solved ... simpler) are also creating this over-engineered feeling.

Writing the above I realised that maybe the specs we write, and the process behind them is over-engineered and this reflects onto how other people perceive the Semantic Web.

Hi Björn,

One motivating use case of how the semantic web can provide added value over existing techniques helps me to explain the concept of meaning over raw information: Imagine you go to Amazon and purchase an introductory book on, say, watercolour painting. Amazon currently can offer recommendations based on what others who purchased that book also owned, reviewed, bought etc. The problem is that Amazon has no idea what the book is about. All it can do is correlate purchasing behaviour on its site.

I'm a programmer, so I buy a lot of programming language and API manuals. Every time I do that, I get bombarded with Star Trek and Buffy recommendation, despite the fact that I am not a huge fan of either. That's because there seems to be a geek purchasing pattern there. It makes a lot of sense to Amazon to exploit that pattern, but what I want is for Amazon to say 'you bought an introductory level book on watercolour painting. We also have this book which is for intermediate level painters and has garnered good reviews."

The UI appearance might be indistinguishable, but under the hood you need something with a much better understanding of the subject matter of books, people's motivation in purchasing certain kinds of books, and what people expect when they start reading a sequence of non-fiction books. You also need something that can reason about that data and apply what it knows to the data at hand. Data mining is not really able to do that, so it is also less able to offer closely tailored recommendations that disregard external factors.

The same concept can be extended to other application areas. Service selection, or service location would require similar reasoning. We are only just beginning to get an inkling of how programming techniques will change, and what capabilities we'll have when we can bring to bear knowledge. As time goes by, people will be able to be more specific in answering a question like this as more real-world case studies are released to highlight the benefits accrued through use of semantic web technologies.

The Semantic Web is a collection of technologies which aim to make a number of data representation and query problems much more trivial.

A hypothetical use case which it would solve better than existing technology such as Google is:

"Show me all available Flights to Location A between X and Y for < $N with a Hotel for $y within 25km"

Though you could answer a question like this using existing technology (xml, sql, relational databases, a programming language), trying to get all of the Hotels, Airports, Travel Agencies and Prices into one giant database would be a nightmare.

Traditional technologies / path to solution: - Represent data in XML - Write guides on parsing it and mapping it to your internal data structures - Sell all of the companies on representing their data in your Custom Format XML - Map XML to a relational database - Build an application which presents views / search inputs which are powered by SQL. - Solve the problem of keeping it all in sync. - Fix problems with duplicate data

Semantic web technologies provide: - A common data representation format, which has structure / meaning (RDF / triples) - Vocabularies to describe that data (OWL/RDFS) - A query language (SPARQL) - Identifiers (URIs) - Data Transport (HTTP)

This means that your Semantic Web path to a solution is: - Find/Develop an appropriate vocabulary (OWL) - Sell companies on modelling their data in it - Use HTTP to fetch data and put it into a triplestore - Build an application which uses SPARQL to query data and present it to the user.

Difficult? Yes, still. Easier? In the long run, for web scale problems, most certainly.

There are surely two aspects to this whole question. On the one hand, your top-to-bottom semantic web application, where instance data (AKA Individuals) is part of the solution, and you have a knowledge base which can be searched semantically. That's one useful thing.

Separately from that, there's also the benefit of introducing semantics into the process of developing data-intensive computer applications. That is, defining meanings for data elements in a way that is independent of the technical design of a data model, and that is comprehensible to and reviewable by business stakeholders and subject matter experts. This is really the application of conventional requirements management principles to data model development, and may be followed by data model design in conventional SQL formats, or by the extension of the model into a triple-store based application as described in the first case above.

There are benefits to the second approach wherever complex supply chains of information are difficult to integrate or costly when they go wrong, for example financial securities information.

I see the semantic web as embedding data within web pages so that you can do useful things with it.

last.fm embeds calendar data in event pages so that tools like the Operator Firefox plug-in can use it to create events in your web calendar.

XFN adds data to web links so that a spidering application can imply social network-type information from them. Some sites, e.g. identi.ca, expose FOAF data that can also be used to deduce the social networks.

Sites like Facebook and Twitter tend to lock the useful data away so that only they, and approved applications, can use it. Data may be exposed in an API, but that requires specific coding. If the data is in a standard format then an application can extract it from any site that implements it.

I think we're still waiting for the 'killer app' to sell the idea to the general user.