Blog Post

Sometimes It's Just Semantics

Stay on Top of Enterprise Technology Trends

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

As someone whose job involves understanding how certain people and things relate to one another, the idea of the semantic web is both compelling and scary. It could make my job that much easier, or it could make me as redundant as switchboard operators are today.

Coding information in a standard way so that machines can see how one person relates to another, or how a string of words could alternately be a movie or a book title, is a challenge. But plenty of companies are taking little bits and pieces of the problem and solving them. One such startup, Radar Networks, the maker of Twine, today received $13 million in funding from Velocity Capital, Vulcan Capital and DFJ. Other startups such as EVRI and Freebase have also benefited from VC interest in the semantic web.

Some of the companies are following the standards offered by the W3C, which is pushing RDF as a standard data structure to underlie the semantic web. But not all companies working on helping machines figure out the relationships and categories that most humans have learned use that standard.

Nor are all the companies interested in making the semantic web work startups. Yahoo uses RDF in some of its offerings and Google‘s efforts with its social graph API initiative resembles the semantic web in its goals. Instead of using RDF, however, it’s using XFM and FOAF tags.

Reuters is another company that sees potential is getting machines to understand relationships. Earlier this month its CEO laid out a pretty compelling vision (at least to Tim O’Reilly) about how Reuters would rely less on delivering information and more on packaging its information in a way that could be used by analysts and computers to quickly delineate relationships.

Reuters would then be able to take its content, make it programmable and offer that data to users, who could then do with it what they will. Things like making relationship charts that currently can take a journalist and graphics department a couple of days to complete, and must then be monitored and changed manually, become easy.

The effort to render all of the data on the web into a semantic form will take a while. Nova Spivack, CEO of Radar Networks, believes that semantic web applications are currently in the early adopter phase. Twine will unveil its efforts in March through a private beta and another startup, AdaptiveBlue launched a semantic plug-in called Blue Organizer earlier this month. Spivack believes that in 2010 mass adoption will take place as people start to expect machines to make “intelligent” connections between people and things.

All of this is interesting, but putting a layer of semantic code over the existing web raises some concerns. One is the danger of inaccurate or at the very least less nuanced sense of relationships between people. Another is the everlasting nature of information on the web. How will coded tags be able to follow the intricacies of human relationships as fights ensue, jobs shift and even names change?

Another issue that we’ll have to deal with is confusion as people try to figure out what the semantic web really is. I’m thinking of it as code added to existing and new web content that helps determine and maybe track relationships between people and contexts for objects. I’m not married to the W3C standards, however, and others are doing this without using those particular programming tools.

There are also plenty of other definitions and hopes for the next phase of the web that may play out before we get an intelligent Internet. It’s already apparent that the web will continue to become more useful over time, but won’t ever replace the benefits of human interactions. If you doubt me, just recall your most fulfilling customer service call with a person compared with your most fulfilling experience with an automated agent. While both are helpful, sometimes you need a real, live human being.

13 Responses to “Sometimes It's Just Semantics”

  1. @ Stacey
    “How will coded tags be able to follow the intricacies of human relationships as fights ensue, jobs shift and even names change?”

    You answered that yourself. You just have to look at it a little different.
    “But not all companies working on helping machines figure out the relationships and categories that most humans have learned use that standard.”

    Humans learn. There is no reason why a machine can’t do that.

    We can for example teach the concepts of “all” or “mean”(meaning “being mean” or “by mean I mean xyz”). Following changing relationships over time is really the easy part.

    Oh and I think the [Ss]emantic [Ww]eb is really so 20th century. Why should I define something a machine can learn all by itself.

  2. Stacey,

    BlueOrganizer has been on a market for 1.5 years. We just launched a new version, called Indigo, which features recognition of things in pages, links and text.

    The key question to me is what exactly is it that semantics helps you do better. For a long time, people thought it would be search, but we no longer think that search is a killer app.

    Semantics helps you understand user context and shortcut search all together.


  3. The final sentence in your article said:

    “…with an automated agent. While both are helpful, sometimes you need a real, live human being.”

    ChaCha has real live human beings answering questions right now. I have had so-so experiences with questions on their web site. But I tried their mobile text service about a month ago and have been using it ever since. It’s great that all these companies are working out semantics schemas but will they ever be as good a human beings…probably not. The big question is can ChaCha make it work financially with people. Don’t know, but maybe. I have 242242 as a contact on my phone because the human at the other end answers my question with no computer ambiguity.

  4. Stacey Higginbotham

    Deepak, I like your thoughts, and assume the people finding people is the common example because it’s easy to understand. Folks have explained the semantic web as the next generation of the back end of the Internet, but when trying to figure out what this means, I’m always given the relationships example. I’d love more examples, so please fire away.

  5. @Alan Wilensky: how about all the thousands of experienced computer scientists who have so far contributed to the development of the standards that underpin the Semantic Web, and the tools that implement these standards? Blood, sweat and tears may have been shed along the way, but I don’t see anyone crying now as the Semantic Web gets ready for the big time. Perceiving the Semantic Web as a top-down approach is one of the oldest and clumsiest misconceptions. You might want to have a read of

  6. The mechanism of the semantic web is fascinating, and for scholastic work it could be a great research boon, but until people learn how to communicate effectively, the semantic web is going to be useless for the masses. As the news media continues to shrink and commercial communication is, more and more, turned over to people who speak in cliches, the semantic web will be of no help.

    A new car is wonderful, but you still have to have a qualified driver to make it go.

  7. in answer to the question: How to deal with is confusion as people try to figure out what the semantic web really is?

    The semantic web is just better applications to the end users. The technology is very much server side in comparison with the glitz of web 2.0 sites. Using web2.0 style sites the suer will say to themselves ” this is cool” that is when the Semantic web will get mainstream appeal. The use of different standrads is worryiny.

  8. FOAF is all RDF, so Google is using RDF as well.

    I admit that I am very fond of RDF, but as a life scientist, metadata means more to me than to most people. I also love microformats and in the general run of things they are very useful. What I find somewhat frustrating is this whole focus on finding people, which is one application, but probably the least interesting. Semantic Web technologies are backend technologies, and to an extent users should be oblivious to them, e.g. it is entirely possible to store relationships between tags on in a RDF store. The point is that the challenge is for developers to take advantage of the backend to serve up more useful applications to end-users, who really should not have to care about what’s going on at the back.

    To quote Jon Udell (who extended a quote by Jeff Jonas); Data finds the data, then people find people.

    What the semantic web allows us to do is make it easier for the data to find the data. Then it’s up to developers to leverage that into making it easier for people to find people.