Monday, 7 June 2010

Semantic Web - a tech fix for a human problem?


I tweeted about the Semantic Web today:
A Semantic Web pursued by techies is futile. How about a Semantic World for starters?
In a then following Twitter conversation with Peter Evans Greenwood, it became clear that we could "dance around" the definition of semantic itself - so how could we ever achieve a Semantic Web?

The real question is: should we even?

I have been working in IT for over a dozen years now, and of all the awkward things I've seen, the -what I call- IT will fix it-approach baffles me most

This approach is common throughout all businesses and industries, levels of organisation, and alike among customers, suppliers and vendors. One could also call it The Law of Self-deceit, or Leverage of Change by Misplacing Responsibility

In simple terms, it means that business and orgsanisational problems that arise to utterly complicated issues, are simply solved by throwing tech at it
  • We have integration issues - let's buy an ESB
  • We have too much legacy - oh we must just SOA it all
  • Our processes are too complex to manage - hmm we need BPM
  • Our business is getting too dynamic and real-time - ah we must have CEP
  • Our users never get out of their application so it's tough monetizing them - let's just level the surrounding walls then and bring the mountain to Moses
  • We don't understand eachother - let's build a Semantic Web
Does that ever work? No, hardly ever. Where we prudent IT guys used to start off with a small pilot before engaging in enterprise-wide battles, nowadays there are "orders from Headquarters" and off we go, like lemmings into the sea; resistance is futile.
However, as a Leverage of Change by Misplacing Responsibility it does function by having people without much subjectivity or personal interests regarding the case in particular (IT people), disrupt the status quo (business and organisational structures) by at least introducing "competition" or a mild threat in the form of another (possible) solution. As such, it always leads to something (usually, revenue for the vendor or system integrator)

However, there is a very bad side-effect: according to my Law of Association, minor or major failures or disasters regarding technologies will hardly ever be (permanently) associated with the circumstances at hand such as people and politics, but only with the technologies themselves. Hence, one should be very careful to abuse technology: your organisation might be very much against it by the time you do need it, and are actually ready for it

If it's too complicated for humans, machines can't do it for you. If it's too simple, boring and massive for humans, machines should (and can) do the job. In between is a gray area of complexity where machines can assist humans

Quoting Peter:
objective semantics are an unreachable goal. the best we can hope for is for dialog to bring us closer
Exactly my point

So, Tim Berners-Lee and others: stop talking and broadcasting about the Semantic Web, just start a pilot and prove its feasibility. Then we'll talk (back)

12 reacties:

Patrick Brinksma said...

It shouldn't surprise you that I fully agree with you. The Semantic Web should be translated to: "What real meaning has the Web", instead of trying to give meaning to the subjective content and language.

Martijn Linssen said...

Thanks Patrick, agreeing with you too on that

I think the Web has grown into something nobody could ever imagine, and will continue to do so. Meanwhile, people are still eager to buy the Emperor's Clothes - mmm on second thoughts, that would have made for a great title ;-)

Tim Molendijk said...

The problem with discussions like these is that Semantic Web is not a technology, but rather a vision for the future of the web.

There is a plenitude of technologies that (can) go by the name "semantic web technology," but this is not a homogeneous group and constantly evolving.

As such, it is hard to talk about the Semantic Web as a technological solution that can be prototyped. Semantic web is not a destination, it's a direction.

Martijn Linssen said...

Thanks Tim!

http://en.wikipedia.org/wiki/Semantic_Web thinks differently though: all tech

A vision for the future you say, a direction: what would that be then, according to you? And why not start with a semantic intranet somewhere? Do we really have to think this big? Aren't we skipping a gazillion of in-between steps?

I mean it's 1999 that the Semantic Web was invented / named, together with XML a.o. Shouldn't we have a little Semantic Something now somewhere?

Tim Molendijk said...
This comment has been removed by the author.
Tim Molendijk said...

Semantic Web in its essence is about increased interoperability and data expressiveness and making the transition from a "web of documents" to a "web of data."

Over the course of the past decade many technologies have been developed to support these ideas. My point is that we must not confuse the technologies with the ideas. If Semantic Web technologies fail (which has happened in the past and will happen in the future) this does not mean that the Semantic Web ideas have failed. New technologies will evolve and new opportunities will arise. (Cf. RDFa instead of GRDDL.)

I agree with you that it is hard to perceive or define what exactly all this Semantic Web thinking has brought us so far. There are definitely some Semantic Somethings emerging here and there, but it's all coming along slower than many Semantic Web proponents wished for.

My personal view is that as long as the ideas of interoperability and data expressiveness are considered Good Things, we will make progress in that direction. Nobody knows which technologies will play a role and what the process will look like, but my guess is we will still call it Semantic Web.

Patrick Brinksma said...

In order for technology to understand human produced content humans have to understand humans a little bit more without the instant judgement layer.

I do agree that the idea should stay alive, as without ideas (and thus imagination) we would have nowhere to go.

Peter Evans-Greenwood said...

Hi Martijn,

I think Semantic Web is the poster child for this problem. For ~10 years it's been a solution in search of a problem, and has failed on all counts. If the technology was so useful then we would have seen significant adoption in this time.

Now we're seeing a wave of renaming — what was Semantic Web is now Linked Data, and is soon to become Web Science — as the cheerleaders try and shift the goal posts. "That thing over there failed, and not my pet project! Viva la revolution!" If we make the term general enough then I'm sure we'll find some success somewhere which, no matter how tenuously related, will be claimed as a Semantic Web success and used to justify all the wasted resources.

I'm with you: lets stop talking about fluffy ideas, silver bullets and peyote inspired visions. Get out there and engage with the problems we're seeing in the real world. Make a difference, rather than chasing a research grant or pushing your favourite technology. Result are what counts.

Martijn Linssen said...

Excellent Peter!

I'm with you all the way, what a great comment!

Makes me think of "And they'll invent Social Media 2.0 or 3.0 which is exactly the same as Social Media now, but just adapted to the time's spirit - leaving the old definition untouched and infallible." which I stated in my Law of Infallibility

Stefan Dreverman said...

First: Tech fix for a human problem. Yes! So is the car, the washing machine, the computer itself, etc... In fact, we invent all kinds of (tech) things because we think we have a problem. It's in our nature.

Semantic web technology is about answering questions, about giving the (web)user the answer he/she needs (or wants). Objective semantics are an unachievable goal indeed. And that's why Tim, in his vision for a semantic web, stated that everyone can determine the semantics of anything on the web.
Where nowadays the owner of a dataset determines what something is or how it is used, Tim wants to put that decision in the hands of everyone.
It shouldn't be the google's, facebooks, shops, ... well companies in general that make data make sense. It should be us.
So, time to step away from huge silos (index) of data (like Google, Bing, Facebook) that can only point you in the right direction. If I want to know how many paintings "Vincent van Gogh", the answer is a number, not millions of suggestions on a search engine result page.
Tim's vision: Make the web itself one huge dataset. So, if we have a question we don't turn to a big indexing server (based on a 1960's approach to information: Catalog, indexes, keywords, etc...) but to the web itself. We ask the question "How many paintings did Vincent van Gogh paint?" and out comes the number. Unfortunately, I cannot provide you with that number yet. :-)

If you want to take a look at some successful projects using semantic web technology:
- www.yummly.com
- www.sindice.com
Apart from the public web, this technology can also be used within organizations. Personally, I've opened the minds (and information silos) of two Enterprise Architecture and Software Development departments by using this technology to store architecture information.

Martijn Linssen said...

Thanks Stefan,

but you give examples of machinery that exactly mimics human tasks at a much higher speed - that is very far from global semantics

I'm all in for global agreement, but stepping away from existing, huge, free solutions provided by big organisations is just hubris brought forth by an urge to prove oneself

The web *is* one huge dataset already. It is its size and dynamics that makes it so complex. Technology is never going to solve that: only mutual agreements are

Stefan Dreverman said...

If you look at the sindice.com example, it searches for semantic data on the web. That is not something that mimics human tasks. It gives to-the-point results with the ability to semantically navigate the web. (Try "Fiat 500" and open the first result)

True, the web is one dataset. But it is not machine-readable. Tim's vision is about making all content universally accessible and machine-readable. In the process of achieving that goal, mutual agreements will/must be made by the keepers of those datasets:

One mutual agreement between two parties only draws one bridge between two datasets. One mutual agreement on the technique used draws an infinite number of bridges between an infinite number of datasets.

Post a Comment

Thank you for sharing your thoughts! Copy your comment before signing in...