Semantic appsThe Semantic Technology Value Chain

September 23rd, 2008 by Greg Boutin

With the slide below on the Semantic Technology Value Chain, I attempted to present my view of the Semantic Web world(s), to inform the discussion recently started through the latest Semantic Web Gang podcast on the top plays in this space. Mills Davis of the Semantic Exchange is working on a much more detailed version of this market segmentation effort, so this is not intended to represent more than a personal overview.

(click on the picture for a larger view)

Looking at this value chain, my working hypothesis is that the initial semweb killer app is likely to be a synergistic Semantic Data Store fed by a popular Semantic App or Semantic Engine. My rationale is that the Semantic Data Store is where most of the semantic web intelligence will reside (dynamically), massive human input is needed to build that repository, and humans won’t put in the effort unless they personally derive some value from doing so. So the proposed recipe is to build a useful app (or more than one), collect and process all the human input collected, do this in a quick iterative fashion, and give third-party programmers some (restricted) API access to your metadata repository as a way to wet the appetites and keep feeding the metadata beast.

Twine and Open Calais are both interesting contenders, but the competition for the title of semantic champion still suffer from a certain lack of heavyweights, although the game is definitely heating up. Only one thing is sure, the value of metadata is on the rise.

Reblog this post [with Zemanta]

  • Jonathan

    Great post! I thought it also might be interesting to consider the last leg in the value chain as ‘semantic data’ vs ‘semantic pages’.
    For instance, I imagine that these class of apps will be working together to annotate a single page (like how you have Zemanta and other plugins on your wordpress page), or be passing relevancy data (like email tags or song recommendations) directly to desktop/mobile apps through apis.

  • Greg Boutin

    Hi Jonathan,
    I like the suggestion even if I’m not sure I completely understand it :)
    What do you envision when thinking of “semantic data”? Are you saying that semantic pages is a little restrictive, and really we should think of the semantic content generated as data rather than pages? If that’s the case, I agree. I used “Pages” to make it very concrete and differentiate it from “metadata”, which is also data. I won’t use “Data” due to that limitation, but I might find a better term than Pages, or change it to Content, which I think might be the best compromise.
    If I missed the point, kindly go out of your way to let me know and clarify, as I think you hit on something important.

  • Jonathan

    Greg, sorry for the brevity, but I think you’ve interpreted the thought as I intended.
    “we should think of the semantic content generated as data rather than pages..”
    2 reasons for this. First, I (and I think others) have had some success in explaining the semantic web concepts to others as ‘finding data was first about relating pages together, links and pagerank. Semantics is about relating data to other data’ – so that is a decent mental model for positioning other ideas. Second, I think that many valuable semantic applications will work ubiquitously, and maybe not interface with the user directly through a browser. For instance, if I had a semantic application to automatically tag my emails, it would not really generate pages, just be enhancing meta-data.
    I like your potential word choice of ‘Content’.

  • Greg Boutin

    I’ve made the change to Content :)
    Couple of minor comments on other points you raised:
    * “Semantics is about relating data to other data” This is a good example that in practice, the semantic web is more and more viewed as a synonym for Linked Data. I wrote about this in my last post on OpenCalais Semantic Proxy. I think this is a mistake marketing-wise. The idea of Semantics imply something completely different, and much smarter, than the concept of linking data automatically. So the challenge will be to explain that to users: “we called it Semantic Web, but actually what it does is primarily weaving data together by recognizing simple entities like dates, people and places”…
    *”If I had a semantic application to automatically tag my emails, it would not really generate pages, just be enhancing meta-data”
    This would fall in the domain of Semantic Application, not Semantic Content. Just to clarify for everyone, I did not imply that the only product of the semantic web are semantic pages, nor that semantic applications only are there to create semantic content. This said, that raises an excellent point: semantic content is NOT the only benefit of the semweb. So how do we categorize and call other benefits? Semantic organization of data? Is that comprehensive enough? Will think about it, and welcome comments.