Tim Berners-Lee continues to preach the gospel of the semantic web. Speaking at the Fourth Annual Bio-IT World Conference and Expo in Boston, Berners-Lee discussed how the semantic web could solve problems in the life sciences:
Life scientists in particular could find the Semantic Web a useful tool and in so doing "provide leadership to lots of other fields" in implementing this next-generation Web technology, Berners-Lee said. "At the moment, I see a huge amount of energy from people in life sciences getting excited by the Semantic Web and what it can do to solve the big-idea problems."
Meanwhile, Clay Shirky has recently published Ontology is Overrated: Categories, Links, and Tags, an essay based on two talks he gave at the O'Reilly Emerging Technology conference earlier this year. Shirky's thesis is that "many of the ways we're attempting to apply categorization to the electronic world are actually a bad fit, because we've adopted habits of mind that are left over from earlier strategies."
In some ways, Berners-Lee and Shirky have the same thesis, but they take off in opposite directions. To Berners-Lee we don't have enough classification and to Shirky we have too much, or at least too much of the wrong kind.
Shirky shows a recent history of classification that goes from hierarchical (filesystems) to hierarchical with some links to hierarchical with lots of links to just lots of links. Ironically, its Berners-Lee's invention, the WWW, that allowed this evolution to take place. As an aside, Tiger's new Spotlight feature promises to do similar things on the computer. I've almost completely given up navigating folders hierarchically anymore. I just use Spotlight to find whatever I need on my computer in the same way I use Google to find whatever I need on the Web.
To be fair, the semantic web is not about a return to hierarchy, but it does ask that things be classified much more actively that anyone presently does. The semantic web is based on ontologies. Shirky gives some characteristics of places where ontologies work well:
- Small corpus
- Formal categories
- Stable entities
- Restricted entities
- Clear edges