In a world where Facebook's oddly named "Graph" API does not allows an authorized user to see friends of his friends, Semantic Web is just a day dream.
You have a good point, but we're not using the web today in the document oriented way of the early 90's.
Sure, if someone will invent a way to use the semantic data in a way that could generate income or market value for the data owner/publisher, then the SemWeb will take off.
However in the current presented form, I don't see a strong incentive for any profit oriented agent to just put out their data in this way.
Depends on which data you're talking about and what the purpose of sharing it is. If it's a catalog of items for sale on an e-commerce site, then sharing that data widely - in order to make those items more discoverable - might be just what a company wants.
True. This means that the result will be just improved vertical searches of products which is kind of what price comparator sites already do pretty well these days.
> Very broadly, things on the Internet will be described with descriptor languages
No. This will not happen.
Look, before Google, internet search relied on people adding meta tags to their pages so that the search engines could know what a page was about. You had two languages, one for humans, and one for computers.
But people don't really care about having correct meta tags or headers or other invisible markers containing instructions for search engines, which is why Google steamrolled the market when it appeared. Google ignored all the instructions and instead analyzed the human language on web pages, and inferred relevance based on that.
If semantic search engines or analyzers or other technology wants to become widespread, it cannot rely on extra markup, it cannot rely on people adding descriptors of meaning to the data it indexes, it has to determine that meaning from the data directly.
Google ignored all the instructions and instead analyzed the human language on web pages, and inferred relevance based on that.
Google are now at least partially on the Semantic Web bandwagon. They, as well as Yahoo, have been detecting, parsing, and using RDFa and microformats data - when present - for a while now, and using it to provide enhanced search results. See:
In terms of people writing metadata into pages by hand, then yes, I agree. But in terms of programatically generating this stuff from data that's already in a database or what-have-you, then I think it will see growing adoption.
The important thing here is to not commit the fallacy of the excluded middle... There's a place for the Semantic Web between "it's a total pipe dream and no part of it will ever come to fruition" and "it will be fully realized in exactly every detail as originally conceived by TBL."
People are already using Semantic Web technologies to some extent, the open question now is "just how prevalent will this stuff eventually become?"
It's not about hand annotating; it's about exposing databases in a well defined way and using standard methods to describe structured data. Look at what dbpedia is able to do with wikipedia data.
Thats a pretty myopic view of data annotation. There is a place on the web, and in the marketplace for data annotated through markup and through ontology, just like there is a need for automated annotation. Its all about the ROI of encoding the annotation.
Granted, I'm not sure what role the semantic web, as defined, will play in this over the long term. And, to achieve optimal costs, I would tend to agree that the great majority of data will need to be annotated by machines.
However, I think it is important to point out that human generated annotations like markup and ontology can play an important role in 2 scenarios: 1) places where there is not enough data to make meaningful machine inferences, 2) vertical domains that have very well-defined structure where the cost of human annotation is actually less than machine annotation.
It'll be interesting to see whether this is widely adopted. As others have said, I've been reading this sort of article for at least 4 years now and there doesn't seem to be much progress.
Why after seven years? From the time TBL invented the original web to the time it saw massive adoption was a decent amount of time, depending on exactly how you want to define "massive adoption." And the problems that the SemWeb are trying to solve are arguably harder, so it makes sense that the ramp up would be slower.
And never mind the fact that there is real world adoption. Google and Yahoo both embraced RDFa a couple of years ago, and have you checked LinkedData.org lately? There's a constantly growing body of data out there in semantically interoperable formats: http://linkeddata.org/
That doesn't contradict the notion that the Semantic Web is of growing importance in any way. RDF triples can be encoded in JSON just as well as they can be encoded in RDF/XML or Turtle or N3 or what-have-you[1][2]. And work continues on a spec to formalize the relationship between RDFa and HTML5[3], so even though XHTML2 got shit-canned, it won't hinder the ability to embed RDFa.
I've tried several times to get really excited about the semantic web, but in the end I tend to arrive at the same conclusion: so far it's still feels like over engineering a solution to a problem I'm not sure we have. A perfect example of this is OWL, which in its pure form is so expressive that it is completely useless for automated reasoning (due to computational intractability). If you were to build an automated reasoning system to solve a real problem you had, you would never arrive at OWL.
Additionally there seems to be a whole lot of reinventing the wheel. The best semweb people are aware of all the past research in logic programming and automated reasoning, but most semweb enthusiasts seem to be hardly aware of prolog let alone that rdf triples are just another way to express what clauses do in prolog.
If we're really trying to solve the 'problem' that semweb addresses we'd be seeing more articles titled "intro to logic programming, knowledge representation and automated reasoning"
I'm with henrikschroder and call bullshit on semantic web being a big thing, let alone the next big thing. It's an attractive idea, but flawed. There are some niches where structured semantic data will flourish (see examples in other comments), but it will be a vast subset of the web.
A better idea is extracting latent semantic information from the existing messy web. However, "meaning" is extremely difficult to characterize, and attempting to encode it in an interoperable manner inevitably leads to a lowest common denominator approach. That will probably still provide tons of value and be much more ubiquitous than structured data, but will ultimately be shallow and fall far short of the vision most semantic web proponents evangelize.
25 comments
[ 3.2 ms ] story [ 28.8 ms ] threadhttp://data.gov.uk/apps - same thing here, driven by the semantic web and linked data.
It's only one small part of the overall Semantic Web, but it's an element that's here and being actively used already.
"It doesn't take economics into consideration. most people/companies aren't that interested in sharing their documents in an anonymous way.
WWW is a web of documents. SemWeb is a graph of data.
More reading: http://www.w3.org/People/Berners-Lee/ http://www.w3.org/DesignIssues/
I think this is a better intro for technical folk.
No. This will not happen.
Look, before Google, internet search relied on people adding meta tags to their pages so that the search engines could know what a page was about. You had two languages, one for humans, and one for computers.
But people don't really care about having correct meta tags or headers or other invisible markers containing instructions for search engines, which is why Google steamrolled the market when it appeared. Google ignored all the instructions and instead analyzed the human language on web pages, and inferred relevance based on that.
If semantic search engines or analyzers or other technology wants to become widespread, it cannot rely on extra markup, it cannot rely on people adding descriptors of meaning to the data it indexes, it has to determine that meaning from the data directly.
Google are now at least partially on the Semantic Web bandwagon. They, as well as Yahoo, have been detecting, parsing, and using RDFa and microformats data - when present - for a while now, and using it to provide enhanced search results. See:
http://www.google.com/support/webmasters/bin/answer.py?hl=en...
People don't give a shit about metadata.
The important thing here is to not commit the fallacy of the excluded middle... There's a place for the Semantic Web between "it's a total pipe dream and no part of it will ever come to fruition" and "it will be fully realized in exactly every detail as originally conceived by TBL."
People are already using Semantic Web technologies to some extent, the open question now is "just how prevalent will this stuff eventually become?"
Granted, I'm not sure what role the semantic web, as defined, will play in this over the long term. And, to achieve optimal costs, I would tend to agree that the great majority of data will need to be annotated by machines.
However, I think it is important to point out that human generated annotations like markup and ontology can play an important role in 2 scenarios: 1) places where there is not enough data to make meaningful machine inferences, 2) vertical domains that have very well-defined structure where the cost of human annotation is actually less than machine annotation.
Perhaps the metacrap rant (http://www.well.com/~doctorow/metacrap.htm) was right?
Seems that way at the moment.
It's seven years later, nothing happened in terms of real-world adoption, it's safe to declare it DOA. In fact it's been safe for several years.
And never mind the fact that there is real world adoption. Google and Yahoo both embraced RDFa a couple of years ago, and have you checked LinkedData.org lately? There's a constantly growing body of data out there in semantically interoperable formats: http://linkeddata.org/
[1]: http://webbackplane.com/mark-birbeck/blog/2009/04/20/rdfj-se...
[2]: http://n2.talis.com/wiki/RDF_JSON_Specification
[3]: http://dev.w3.org/html5/rdfa/rdfa-module.html
Additionally there seems to be a whole lot of reinventing the wheel. The best semweb people are aware of all the past research in logic programming and automated reasoning, but most semweb enthusiasts seem to be hardly aware of prolog let alone that rdf triples are just another way to express what clauses do in prolog.
If we're really trying to solve the 'problem' that semweb addresses we'd be seeing more articles titled "intro to logic programming, knowledge representation and automated reasoning"
A better idea is extracting latent semantic information from the existing messy web. However, "meaning" is extremely difficult to characterize, and attempting to encode it in an interoperable manner inevitably leads to a lowest common denominator approach. That will probably still provide tons of value and be much more ubiquitous than structured data, but will ultimately be shallow and fall far short of the vision most semantic web proponents evangelize.