Wednesday, May 09, 2007

Social and Collaborative Construction of Structured Knowledge

This was the title of one of the workshops I attended.

I was pretty shocked by the contrast between the academic and the commercial worlds (the latter was really under-represented in this meeting) on their views of how to collaborate to construct structured knowledge.

Most if not all of the solutions presented on this workshop were tools designed to have a group of scientists to collaborate on building ontologies. Why scientists? basically because to build an ontology which is the agreed form of storing structured knowledge you pretty much have to be a geek or an engineer of some sort.

Jamie Taylor from MetaWeb was my favorite talk. Their goal is rather ambitious: Represent all human knowledge. They are using some type of proprietary graph based on ontologies which they totally hide from the user through a very nice ajaxed interface. The end user can define the objects and properties they want extending on top of a few basic types: int, string, language string, date, object. I'm hoping to get my invitation to test soon as I don't totally understand who has the right to modify what yet, apparently they have some sort of sandbox, but not sure how the procedure is.

Anyway, the talk was very interesting in itself because of the following facts:

1. He used the story of "the stone soup" to make his point about how it's all about multiplying a scarce resource through the community (in this case data).

2. There were plenty of references to interesting concepts or facts.
. One of them was Bradley Horowitz's post stating that the collaborative websites have three types of users: creators, synthesizers and users pointing at how loyalty should be developed towards creators but also the focus should be on being a community at all scales to become sustainable.
. Another about Duncan Watt's experiment on Cumulative advantage, stating the inner unpredictability of the winner in a community based world.
. Also, Stuart Kauffman on "The search for the laws of self-organization and complexity" talking about NK networks.
. In addition, Gary Hamel on his book Leading the revolution" stating that "radical innovation is the competitive advantage for the new millennium".

3. He explained their business model which is innovative and interesting. Basically they build Freebase and promise it will be an eternally free source of information. On top of that they created Metaweb which will build and sell services on top of Freebase. Their philosophy seems super open (in reality as they will provide APIs for information retrieval not only for data entry) and they are into the non-zero-sum game which is a definite plus.

4. He was open on how freebase is doing it by:
. Focusing a lot on the User Interface as that's all the users will have. Every assertion missed is lost.
. Planting lots of seed data.
. Using a model of canonized objects to avoid duplicates where there is only one object per person, etc.
. Having a good politic model and being lousy coupled (APIs and UI hiding complexity).

A great fun talk packed with all sorts of information!


Gabriel Medina said...


Do you have one invitation code to freebase for me?.

Cecilia Abadie said...

Sorry Gabriel, they did not open this choice for me to have invitations to give away. If they do you're in my list ;)