Friday, March 28, 2008

Blog links - Not Just Simple Pointers

Links have played a central role ever since Tim Berners-Lee created the world wide web in 1989. Yet his vision hasn't fully materialized. Links between documents are useful, but have no meaning. The semantic web, was intended to put the meaning back into those links so that we could pragmatically know what we are pointing to.

Yet the semantic web hasn't really taken off. The current web has a lot of inertia, and works very well. Getting people to switching over both in thought and practice to a web of data is a difficult task. Instead, I'd like to approach the problem from the opposite direction: Let's not create a whole new web, let's augment the old web. Instead of creating a new structure for the web using triples, let's pragmatically annotate existing links using machine learning, natural language processing, and any other technique we can code up. These power granted to us (and our machines) by these annotations should provide us with the leverage to solve an entire slew of problems.

Ranking the influence and authority of a blog is one such problem. This upcoming week I'll be presenting my work on modeling blog links along 3 dimensions to ICWSM. Each dimension tries to answer a key question. Why did the author make this link? What is the link really pointing at? What does the author feel about the target?

1 comment:

Median Joe said...

Hi Justin,

I think the idea you described is very appropriate. I also believe we should take advantage of any computer-assisted annotation opportunities, and I'll talk about additional ideas for bootstrapping the semantic web in tomorrow's class.