Slashdot Mirror
Why the Semantic Web Will Fail
Jack Action writes "A researcher at Canada's National Research Council has a provocative post on his personal blog predicting that the Semantic Web will fail. The researcher notes the rising problems with Web 2.0 — MySpace blocking outside widgets, Yahoo ending Flickr identities, rumors Google will turn off its search API — and predicts these will also cripple Web 3.0." From the post: "The Semantic Web will never work because it depends on businesses working together, on them cooperating. There is no way they: (1) would agree on web standards (hah!) (2) would adopt a common vocabulary (you don't say) (3) would reliably expose their APIs so anyone could use them (as if)."
Thank God for Web4.1!
One of the problems is lack of standardization, and one of the symptoms is Yahoo! normalizing Flickr's user accounts with its own?
The semantic web will fail because it is too complex and noone outside the academic community working on it really understands it. The ad-hoc tagging systems and microformats Web 2.0 has brought are good enough for most people, and much simpler for the casual web developer to understand.
Doesn't Web 2.0 reach a "critical mass" as some point, where busineese will no longer be able to not cooperate? Of course, it all gets very fragile even then...
...says the guy who's blogging this opinion...
ilovegeorgebush
It was created to solve a problem we had when everyone was using Hotbot and Altavista, but people are trying to introduce it into a world where everyone is using Google. (And Wikipedia. And all that Web 2.0 junk.)
I don't need you to mark "This page is a REVIEW of a CELL PHONE that has the NAME iPhone" anymore. All I need to do is Google "iPhone review" or hop on over to Amazon. Problem pretty freaking solved from my perspective.
Help poke pirates in the eyepatch, arr.
It might fail for the reasons given (no I've not read the full article yet - naturally) but personally I think it will fail simply because it's too much work for the amount of payback. It would be great if one day magically over night all our data was semantically marked up but that's not going to happen. The reality of it is that we will have to mark up the majority of content by hand. Even then inter-ontology mappings are so difficult that I'm not sure the system would be much use.
Perhaps worse than that though is the prospect of semantic spamming. It would be impossible to trust the semantic mark up in a document unless you could actually process the document and understand it. What would be the point in the mark up in that case?
I used to have a better sig but it broke.
So what is this semantic web / web 2.0 thing anyway?
Sure, we're all seeing community sites, blogs, tagging, etc. But each of those sites is an individual site, and their only connections seem to be plain HTML links. Community sites don't really allow collaboration, blogs are standardized personal web pages and who here uses tags to actually find information? All these things might warrant a "Web 1.0 patch 3283" label, but is it really a new type of web? Is it the type and magnitude of paradigm shift that the first web was? It only seems like people are just becoming more aware of the possibilities of the same web it was 10 years ago.
Slashdot social media options: AIM, ICQ, Yahoo, Jabber and Mobile Text. Why no MySpace?
The thing the academics who push the semantic web fail to consider (most of the time) is that the Real World does not function like their Ideal World. In the Ideal World, everybody cooperates and works together to produce something of value for all mankind. So we get lots of correctly and appropriately marked up pages that give useful information on what's stored therein.
But in the Real World, any online system that is used by a large enough number of people will eventually become attractive for spammers and scammers to defile and twist to their own purposes. So you'll get a deluge of pages that appear to be useful reviews of digital cameras (and are marked up as such) but in fact simply go to a useless "search" page that has lots of link farm references.
And if you say "Ok, so we don't trust the author of the page, we have someone else do it"... then who? Who's going to do all the work? Answer: Nobody. AI is nowhere near being smart enough for this. Keyword searching is, unfortunately, here to stay. If you trust the author to do the markup, then the spammers have a field day. If you say "Only trusted authors" then the system will still fail, due to laziness on most people's part - if a system isn't trivial to implement and involves some kind of "authentication" or "authorization" then nobody will use it, period. The Web succeeded in the first place because anybody anywhere could just stick up a Web server and publish pages, and it was immediately visible to the whole world.
The Semantic Web will fail for the same reason that the "meta" tag failed in HTML: Any system that can be abused by spammers, will be abused.
So, the Semantic Web, which is all about helping people find stuff, will fail. Not because of any technological shortcomings (it's all very nice in theory), but simply because we as people won't work together to make it work. Well, a small number of people could work together, but as that number got larger, until it reaches the point of being useful, it will automatically get to the tipping point where it becomes worthwhile for the spammers to jump in and foul it all up.
Lewis Carol had it right, and George Orwell agreed with him: "Which is to be master" is the question that matters.
In free societies, everyone is master, and our language is conditioned only by the minimal need to communicate approximately with others. Beyond that, we are free to impose whatever semantics we want, and we do this to a far greater extent than most people realize. As a friend who works in GIS once said, "If I send out a bunch of geologists to map a site and collate their data at the end of the day, I can tell you who mapped where, but not what anyone mapped." Individual meanings of terms as simple as "granite" or "schist" are sufficiently variable that even extremely concrete tasks are very difficult.
Imposing uniform ontologies on any but the most narrowly defined fields is impossible, and even within those fields nominally standard vocabularies will be used differently by rapidly-dividing "cultural" subgroups within the workers in the field.
The semantic web is doomed to fail because language is far more highly personalized than anyone wants to believe. I think this is a good thing, because the only way to impose standardized meanings on terms would be to impose standardized thinking on people, and if that were possible someone would have done it by now. Whereas we know, despite millennia of attempts, no such standardization is possible, except in very small groups over a very specialized range of concepts.
Blasphemy is a human right. Blasphemophobia kills.
But there are three ways to get that.
1) A search service that indexes all of Romario's goals.
2) A manually built asset that aggregates all of Romario's goals.
3) A standard system of semantic tags that self-identifies all Romario goal assets.
#1 is Google. As you point out now it relies primarily on keywords but you oversell the problem in two ways. First of all most video hosting sites already provide author and/or community tagging--thus providing a way for keywords to be assigned. Second, you're comparing a future semantic Web against the Google of today.
#2 can be provided by commercial video companies now ("1,000 Great Man U Goals," etc). It's also possible that a fan site could do the manual labor to find, upload, and keyword the videos.
#3 is the "semantic Web" approach, wherein all content providers follow a standard for self-identifying their content in a computer-parsable way.
The thing that distinguishes 1 and 2 from 3 is the scope of work required. #1 and #2 rely on a small team of dedicated people to accomplish the task. #3 relies on a very broad group of people of varying levels of dedication.
If you're talking practically about the solution, none of those approaches are going to to get to 100%. As others have pointed out there is a real human semantic problem in identifying which goals of Romario to count, how far back to look, etc.
But the key is that #1 and #2 are approaches of a scope that we know can work. #3 seems unlikely to get the buy-in and effort required.
Build a man a fire, he's warm for one night. Set him on fire, and he's warm for the rest of his life.
One of the features of the W3C's model (based on RDF) is that it doesn't push the idea that everyone should adopt the same vocabulary (or ontology) for a topic or domain. Instead it offers a way to publish vocabularies with some semantics, including how terms in one vocabulary relate to terms in another. In addition, the framework makes it trivial to publish data in which you mix vocabularies, making statements about a person, for example, using terms drawn from FOAF, Dublin Core and others.
The RDF approach was designed with interoperability and extensibility in mind, unlike many other approaches. RDF is showing increasing adoption, showing up in products by Oracle, Adobe and Microsoft, for example.
If this approach doesn't continue to flourish and help realize the envisioned "web of data", and it might not after all, it will have left some key concepts, tested and explored, on the table for the next push. IMHO, the 'semantic web' vision -- a web of data for machines and their users -- is inevitable.