Semantics and the GeoWeb

I went along to a very interesting and well delivered presentation in front of it must be said a rather disappointing audience at the British Computer Society in London yesterday evening. “Make Mashups Correct, Complete, Relevant and Revisited” was a presentation originally given by Jonathan Lowe of Giswebsite LLP at Where 2.0, and as Jonathan is a great presenter I was really looking forward to it.

The presentation actually focused on the currently rather specialised area of semantic spatial databases and their potential in powering the mash-ups of the future. He high-lighted some of the darlings of the semantic database industry freebase and True Knowledge, who have developed technology that really demonstrates well the benefits of semantic databases.

The benefits come from having a much more structured data modelling approach than we have become used to on the web, the demo of freebase here is a great example of this, but such a strongly typed approach is also the major weakness of semantic databases at the moment.
Who defines and categorises data into these types and who builds the relationships between database elements. The wiki approach that freebase uses is a great start but ultimately will it scale ?

Semantic databases will become the future way we interact with information online only when their development and maintenance can become automated, in the same way that the creation and analysis of the web indexes behind web search is automated.

In the meantime that make some great demos

Written and submitted from home, using my home 802.11 network.

4 comments

  1. Jonathan Lowe

    As ever, Ed, your input sparked lively and interesting discussion during the Q&A session. Thanks for turning up on a Monday night and for your kind words about the talk. Later, at the pub, without you there to clarify, several people attempted to interpret your comment about semantic databases’ scaling issues. I understand your observation that it takes manual effort to establish the categories or types initially; I also find that to be the the trickiest step, as different people often want different structures depending on their particular business cases or conceptual frameworks. Once a type is agreed and people start populating a type with topics and properties, however, the bi-directional relationships are automatic (at least, they are in Freebase). For instance, by entering “Christopher Wren” into the “Architect” property of the Royal Observatory topic’s “Structure” type, a “Royal Observatory” value is automatically added to the Christopher Wren topic’s “Structures Designed” list. The idea is that if a small group establishes the initial structure for a body of potentially popular types, then simple entries by the masses are automatically digestible by sophisticated browsing tools like David Huynh’s Parallax (http://mqlx.com/~david/parallax/). You might find the results of a parallax search for “mobile phone” intriguing (see http://mqlx.com/~david/parallax/browse.html?type=%2Fuser%2Frobert%2Fmobile_phones%2Fmobile_phone).

    That said, perhaps you had a different approach to building and maintaining databases in mind that relies on crowd-sourcing techniques? Sounds like an idea ripe for execution by Google. 😉

    Thanks again for attending and for your write-up.

  2. Ed

    Jonathan,

    Thanks for you input, my questions about scalability are as you point out to do with the initial relationship building around any particular subject, but also just the scale of the challenge. If you think about the about of information that exists in an unstructured form already on the web, it will be impossible for people to harvest this and structure it without some form of automation.

    There is a reason we are not using the yahoo catalogue of the web anymore…

    ed

  3. Dan Rickman

    Ed – thanks for the write up and also to Jonathan for a fascinating presentation and discussion. I think the automation issue is important – relates to a basic issue of classification/taxonomies and linguistics which in my view can ever be solved sub-optimally. The other issue of course is crowdsourcing – and I am now reading Clay Sharkey’s book to find out more!

    Speaking as chair of the geospatial SG, yes it would have been great to see more people there and I’d encourage everyone who is interested to come along and take part. We have more events already planned and more in the pipeline…

Post a comment

You may use the following HTML:
<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>