Blog Post

Meet Silk, the Semantic Web for the rest of us

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Almost since Tim Berners-Lee first came up with the original concept of the web, there’s been a new, improved version on the horizon: the much-promised Semantic Web. This, goes the thinking, is a way to make the online world more useful by categorizing everything on a page with a layer of extra information — data that can tell your browser that one particular series of numbers is a date, say, while another is a price. This, in turn, allows your computer to understand more about the information it processes, theoretically making it easier to identify events on the date you’re looking for, or items at the price you’ve chosen.

It’s an ambitious goal that has been chased by many folk over the years — and yet, for all the effort to bring the Semantic Web to life, the reality of the situation is that it always feels just out of reach …  something talked about and hoped for, but never achieved. The truth is that while some progress has been made, we’re still a long way from reaching that vision.

Or are we?

Silk, an application coming out of private beta on Thursday, has a grand ambition to build a version of the Semantic Web that we can all use.

It works like this: you build a series of pages inside Silk and link them together with tags. The pages can be anything you like: text files, pages from your website, company documents, your schedule. And the tags, too, can cover anything you like: any genre of data you can imagine.You then use the simple editor to add tags to your documents (telling it, for example, that the “United States” in your file refers to a country) and it connects the dots for you.

For example: say you’ve got a set of files with information about various countries and you want to rank them by GDP. While traditionally you’d have to do it manually — or at least get an underling to compile a spreadsheet from the data in your documents — Silk makes it far easier to perform the same task without requiring those extra steps.

If your Silk pages are tagged properly, the app simply joins the data together and allows you to instantly build charts, graphs or maps documenting the data you’re after.

Descriptions don’t quite do it justice. To really get an idea of how it works, you have to see it in action.

To me, it’s like some sort of cross between Wolfram Alpha and Wikia, allowing you to perform calculations and answer questions using your own information — and on first glance, it’s really amazing, pulling data together in a fast and powerful way. Imagine if you could perform actions across a huge database like Wikipedia: I know it would save me a huge amount of time.

Something so powerful doesn’t come from nowhere, and it turns out the service is the product of a Dutch team consisting of 11 people, who have been working on the technology for the last couple of years with seed funding from Niklas Zennstrom’s Atomico.

But now things are open to everyone, they could move forward much faster.

Silk’s operations chief, Sander Koppelaar, told me that the public launch was just the latest step in their mission, which is now moving from a lot of behind-the-scenes work with a beta of 10,000 registered users to something more public and open.

“The Semantic Web is a very overwhelming problem, but we’re trying to solve it in a different way, our own way,” he said. “We’ve been developing for over two years, and we’re shifting more and more to user-facing improvements.”

That’s a fair recognition that however useful Silk is right now, it’s still got a long way to go — and in particular there are three issues that I think Silk needs to solve before it can become a significant service.

First, there’s the problem of proper tagging. Without tags, the service is nothing — and tags have to be added manually right now. That’s a laborious job, and the company wants to make it much easier in the future. Koppelaar said that automatic tags on common words, suggestions for which tags may be appropriate and “instant gratification” when something is tagged are all being worked on for future releases. But it’s still a time-consuming roadblock right now.

Second, and somewhat related, there’s the fact that it’s a cloud-based service. That means users can currently only create pages hosted on a subdomain of Silk’s site. This will soon change, he said, with custom domains, extended private use, and company use with larger data sets all probably being paid-for services that will help generate revenues.

Then there’s the fact that each set of Silk documents reside inside a silo, unconnected with any others. While that’s got it uses — not least because your lexicon of semantics may be different than mine — it also means that anyone wanting to use data in Silk must first import it. And if two different people wanted to replicate the Wikipedia example in the video, they would both have to import and tag those pages. Surely it would be much more powerful if there were a common set of data types that could be accessed by all Silk users?

Koppelaar says that there are plenty of additions in the works, and that while it will always remain a cloud-based service, the ability to bring data in through APIs or other widgets is definitely on the cards.

And if those problems get solved? Well, then this could be a really interesting twist for big data and the cloud … and perhaps bring the Semantic Web dream just a step closer to becoming a reality.

20 Responses to “Meet Silk, the Semantic Web for the rest of us”

  1. Reblogged this on Briskin, Cross & Sanford, LLC and commented:
    The Semantic Web is coming! Silk is an interesting step towards the goal of creating context in information floating about the Web. It’s just a step, though and more is coming. This is a concept who’s time has come and its arrival will open up not only new uses of data inhabiting the web, but will make both the underlying data and the uses that can be made of it infinitely more accessible.

  2. Elad Kehat

    I’m sure that the technology is absolutely gorgeous, but Silk had better find some better use cases to promote itself.
    “Which EU prime minister born before 1955 has a background in science”? Come on.
    I’m sure that Silk can come up with questions that people may actually want it to answer for them.

    • Bobbie Johnson

      I certainly think the demo could do with a better example, or at the very least better context for the example.

      What counts as a better example may well depend on who your users are, though. Drilling down into a data set in this way can get very particular, very fast.

  3. Ayn Rand

    Self-organizing of information using semantic indexing is becoming real and popular. I recently downloaded an Android app – My Tags from Zahdoo from Amazon appstore and it has the ability to automatically extract key tags and automatically organizes Reminders, ToDos, News and related items together.

  4. Self-organizing of information using semantic indexing is becoming real and popular. I recently downloaded an Android app – My Tags from Zahdoo from Amaazon appstore and it has the ability to automatically extract key tags and automatically organizes Reminders, ToDos, News and related items together.

    • Bobbie Johnson

      The guy at the top is Tim Berners-Lee, inventor of the web — quite a well-known individual — and the person being talked about at that point in the story.

      The guy at the bottom is Sander Koppelaar, the ops chief at Silk, and the person talking at that part in the story.

      I thought it was easier not to refer to them in the text (you never know whether people can see the images or not) but if you’re in most browsers, you should be able to mouse over the picture and see a caption.

  5. Buster McDermott

    Interesting idea, though the idea of all the contextual data being sort of silo-ed off rather than being shared strikes me as problematic. Real humans monitoring the tagging and keeping things current and accurate will also be totally crucial to making this work, as sites currently utilizing Freebase data (like, for example) demonstrate. The whole system breaks down if the tags are misapplied.

  6. ronald

    I think in the long run we will to get to self organizing data(context), the first step might be a mash up of data finds data and Google Goggles with Siri (the hidden services) combined with and augmented with mobile sensor data(mainly space time).

    The problem is we got a real world and the virtual world in our systems, but we work from a “mental” model which combines both.

    I don’t think partial models which have to be augmented manually will do, or I’m just lazy. Which brings us to the question is the problem augmentation of data or data binding(means we got the data already). Just imagine if Google would wake up one day and combine the above systems and data with flexible data binding in space time, they certainly have the point solutions now they only need the data binding.

    • Thorsten Luedtke

      You don’t get it: TBL is not trying to build a single database but sets distributed ‘databases’ (i.e. structured web sites) authored by independent administrators. If it was just a central database (like Silk, LinkedData, etc. are trying to do) then the problem would be solved by just pointing to the Business Intelligence community: They’ve done it since the 80’s. However, trying to implement TBL’s vision is like asking the United Nations to agree on a common language for everything in the world (they can’t even agree on annihilating child poverty..) and therefore it’s impossible to design it in a ‘top-down’ fashion (and don’t forget that language is a moving target..). So, it has to be done ‘bottum-up’. But: If there’s no common dictionary (i.e. single ontology) then tagging the same things with the same tags is impossible. Therefore the semantic web is impossible. TL

      • Bobbie Johnson

        This is really the key difficulty in making the Semantic Web a reality: it can only work it’s organic… but it cannot work if it’s organic.

        There’s a halfway point, perhaps, where a series of bottom-up ontologies gain enough traction that they can be supported by a few top-down ones — and still give people room to manoeuvre. I’m no SW expert, but it seems we can look to the world of linguistics and see that a lingua franca can emerge naturally without homogenizing *everything*.

      • davidcdean

        Gotcha. Well I hope it works out for them. Every time the topic bubbles up I’ve always pictured something like the cross-linked data browsing described in Snowcrash. I’m not sure how this could ever catch on as it’s implemented here, but I guess we’ll see.