views:

1124

answers:

9

I've heard a lot about the semantic web but I'm still not exactly sure what it is. How will it be different to the web we know now?

+19  A: 

How will it be different to the web we know now?

Right now the HTML+CSS is centered more on structure and presentation. Semantics is about the meaning of the information. In semantic web you use shared ontologies to establish meaning (semantic) of the object and meaning of relations between the objects. Best known ontologies are: FOAF and Dublin Core.

Typically semantics would be expressed in specialized language, such as RDF or OWL. RDF can be embedded within XHTML using eRDF or W3C's RDFa.

Less structured alternative to eRDF/RDFa are microformats.

Read more at: http://en.wikipedia.org/wiki/Semantic_web

vartec
Dublin core relates to what an information resource contains (data about data / metadata), rather than the relationship between information resources.Information about relationships can be described using ISO topic maps (http://en.wikipedia.org/wiki/Topic_Maps).
codeinthehole
As far I know - the semantic web is more involved with the relationship between data (rather than classification).
codeinthehole
+1  A: 

Currently with HTML pages we have markup tags which describe how content should be displayed, <b>, '<pre>, etc. These tags imply no meaning about their content.

The concept of a semantic web is that documents would contain XML tags that do imply meaning about their content. For example <person><firstname>. The grand idea is that CSS would be able to format documents such as these but it would also be possible to extract meaningful info easily from these documents.

AnthonyWJones
That HTML elements have no semantic meaning is a common misconception. HTML covers only the meaning of an element with respect to the structure of a document. HTML has document-focussed semantics not data-focussed semantics.
Jon Cram
It didn't say HTML had no semantic meaning at all, I said that they imply no meaning _about their content_. As you say HTML is structual have meaning towards the renderer. HTML is agnostic to the content itself.
AnthonyWJones
The Semantic Web with upper-case S is RDF. The use of XML is only incidental.
hsivonen
I don't see how changing the case changes the word or that RDF has a monopoly on the concept. XML however is a reasonable basis for delivering the concept regardless of whether you subscribe to a specific implementation of "Semantic".
AnthonyWJones
@AnthonyWJones: Fair point, I slightly misread your answer! Hopefully someone reading these comments will learn something from them.
Jon Cram
A: 

It's a buzz word to attract people's interest, Similiar to Web 2.0

I.e. In the future content will be split from presentation allowing much goodness.
In reality Facts will be subjective, depending on the realibility and authority of the host.

In other words, users won't see much difference from now.

Bravax
The difference is that the computers will know about reliability and authority
Simon Gibbs
I doubt it. Those are both subjective terms.
Bravax
The fact that for example "jaguar" means both an animal, a car and few other thing is subjective? Temperature expressed in Kelvin scale is subjective? What are you talking about?
vartec
No those aren't subjective, but say whether a wikipedia article (on jaguars) is accurate is subjective. In the future we will have some semantic websites but the vast majority won't be.
Bravax
Semantic is about specifying, that when you write eg. jaguar, you refer to animal, not the car. What has it got to do with Wikipedia?
vartec
The wikipedia example is referring to the reliability and authority of the semantic tagging.So I could create a website and specify it relates to vartec's details for example, and machines would be none the wiser.
Bravax
You actually hoped for global AI?
vartec
No (I'd be out of a job.), but that's what some answers imply.
Bravax
I very much hope for global AI. I, too, may be out of a job, but I'd at least get a very long holiday.
Jon Cram
The semantic web allows for tagging trustworthy sources, which will make it different than the web now in that you can delegate the task of source-checking to the computer.
Mark Cidade
What marxidad said.
Simon Gibbs
I can see that working brilliantly theoretically, but being an absolute nightmare in practice. But I know there are better people then me working on this.
Bravax
+11  A: 

The best explanation is by example. Try googling for all cars advertised on the web with engines smaller than 2.0 litres that run unleaded, and have an mp3 connection and can been seen in a showroom conveniently accessible by public transport from my house.

Google just won't be able to help you with that query, not really. You have to make several searches and correlate the results yourself. On the Semantic web, you'd be able to express an interest in products for sale that are cars, and add the constraints. Every result would be useful. One or more UIs might enable you to do that, some may be specialized, others entirely generic.

An other example, creating a chart of things not normally stored in one place, say the popularity of diet coke, or country walks in a population versus the levels of clinical obesity in the same population. For these you may not use a web browser at all, but might use something more like Excel - but the semantic web gives you tools (SPARQL, RDF) for finding and manipulating the data that is out there and is accessible via HTTP.

So the point made by Bravax is not entirely true, not a lot may change - you may merely get some more useful and better mashup web sites. Or you may find yourself doing a whole lot of stuff you never thought of as being related to the web before today.

The current web has lots of alternatives for doing the same thing, say Animated GIFs, Flash, Silverlight, DHTML etc. For putting data on the semantic web there will be a range of tools and formats. RDFa is a good one, a more general type of microformat, but you could provide a dump of the whole database, expose a SPARQL endpoint, use a microformat or a proprietary HTML structure and add a transformation, there will be many tools to suit different cases.

So Vartec is also partially right, you may use RDFa and eRDF but you could also use a whole lot of other things for publishing data.

Note that there is a lot over overlap between the semantic web and another simper concept called Linked Data. How they relate to each other isn't clear, but my perception of it is that the Linked Data web is what you need before Semantic Web tools and techniques have anything to do. Linked Data is about data, the semantic web is more about processing the data, reasoning over it and handling issues like trust reliability and such like. Essentially the bottom few layers of the technology stack.

Simon Gibbs
Good luck with end users entering SPARQL queries into a search engine.
hsivonen
I think users will enter SPARQL about as often as they enter SQL now, though perhaps slightly more often as they'll have more data to attract them. I think you'd be appalled to see how much SQL there is embedded in Excel spreadsheets sometimes!
Simon Gibbs
My search engine example did not mention SPARQL anyway....
Simon Gibbs
"Linked Data" is just a euphamism for Semantic Web
Mark Cidade
Quite possibly, though Linked Data people tend to look an "ontologies" and other things a bit differently. Sem Web people tend to look at things from a maths / logic point of view. Linked Data people the grunt programmer perspective. Compare Open Link Software to Clark and Parsia for a contrast.
Simon Gibbs
As a response to hsivonen's comment: I don't think Simon meant the users would enter SPARQL into a search engine. I interpreted the answer as meaning a user could enter the query in the sentence-format mentioned and the search engine could parse it into SPARQL or something similar and then fetch data based on semantically entered data (on the Web) that could be correlated automatically.
T Pops
Natural language is just a way of expressing criteria, there are other easier ways of expressing criteria. SPARQL was mentioned as a way of getting your data (e.g. the list of cars for sale) onto the web, as an alternative to say the DataRSS or Froogle feed that you might be providing today. There are other roles for SPARQL as well, which are far more interesting than web search backends and these were also mentioned.
Simon Gibbs
+4  A: 

The Semantic Web is at heart a really simple idea. (Like all the good ones.)

The Web at present consists of documents with links between them. Google have made a pretty good business out of using context, and anchor text within the links, to work out what the links mean and build an engine for retrieving data based on that. In other words, Google guesstimate what the semantic meaning of a link is.

The Semantic Web idea is "what if these links were typed?" Every fact on the Web gets an address - a URI - and is linked to other facts (also URIs) by relations (also URIs). Groups of relations are called "ontologies".

So instead of page A links to page B, like on the current web, links on the Semantic Web are more like:

URI A links to URI B with a link of type URI C.

Anything can have a URI. People can have URIs; usually we use a set of relations called FOAF to describe them. So let's say the URI for Jeff Atwood is http://codinghorror.com/foaf.xml; then you could say:

<http://codinghorror.com> <http://xmlns.com/foaf/0.1/homepage> <http://codinghorror.com/foaf.xml>

ie, http://codinghorror.com is the homepage of the person represented by the contents of http://codinghorror.com/foaf.xml.

Now machines can read, and query, these relationships - so you turn the Web into a database that computers can immediately do something with. The Semantic Web query language is SPARQL, and it's worth checking out.

adw
A: 

The Semantic Web is a distributed information system where interlinked data is published as RDF triples over HTTP. RDF triples consist of subject, predicate and object but can have other things attached to them such as datatypes and annotations about the natural language of objects. On the Semantic Web, URIs are used both as identifiers and as addresses of network resources.

It’s different from the Web, because the Web is a distributed information system of documents and application interfaces.

hsivonen
A: 

Semantic web is an utopia, just like SOA. They're both dead before even born.

User
-1: Thanks for trying.
Jim G.
A: 

The Semantic Web is what Tim Berners-Lee, the inventor of the World Wide Web, really intended the Web to be—that is, a global graph of interlinked data. It is a generalization of a social graph, where you can use social data (with vocabularies like FOAF) as well as any other kind of machine-understandable data and connect them to each other. The standard formats for describing this infortmation to machines is the Resource Description Format (RDF) and the Web Ontology Language (OWL). There's already a lot of encoded data on the Web, including an RDF version of Wikipedia, called DBPedia.

The Semantic Web will be different than today's Web in that computers as well as humans will understand what documents contain as well as what the significance of the links between documents are. This will facilitate automation of information-processing tasks, including researching information from trustworthy sources. The full SemWeb stack includes cryptography, proof systems, and trust networks.

Mark Cidade
+2  A: 

The Semantic Web is just that - a Semantic (meaningful) layer on top of the WWW. It is semi structured (RDF), it is self-describing (ontologies using OWL), and allows resource discovery (SPARQL).

The Semantic Web works on the premise of the "Open World" assumption; just because something is not stated doesn't mean it doesn't exist, it is simply "unknown". This is a fundamentally different logic to that used in an RDBMS like MySQL et al. - if something is missing it doesn't exist - "Closed World" assumption. Prolog and DATALOG are good examples of Close World logics.

If you want to really learn what is happening underneath, you'll need to look at its foundations, which lie in Description Logic. A good overview of the Description Logic can be found here: http://www.inf.unibz.it/~franconi/dl/course/

If you want to learn more about RDF, read the RDF Primer. RDF Semantics is another rip-roaring read.

Researchers have basically given up on the "Semantic" part of the Semantic Web and decided to focus on Linked Data - how RDF triples can be navigated so that we can waste more Internet bandwidth ;-)