sardonick

Attended a talk today by Stefan Decker of DERI in Ireland this morning. “Semantic Web 2.0″ was the title of the talk–and I think Stefan wins the Most Buzzwords in a Talk Title award.

I must admit when I came into the talk that I was a bit skeptical–Semantic Web 1.0 never got off the ground (not the way the WWW did, at least!), so are we really ready for a 2.0?

Stefan is of the view that Semantic Web (1.0) never took off at the time because we didn’t have the tools and connectedness for it to reach critical mass of adopters and ease of use. It was like “people dreaming of building a fighter jet when they only have parts to make a bicycle”, as he puts it.

So, I guess with all this Ajax and social networking and folksonomic whatnot, we finally have the tools to help the Semantic Web really succeed? He thinks so. Oh, that I could be so optimistic!

Snarkiness aside, here’s some of the goals he sees within SemWeb 2.0:

• semantic interlikning of online community sites
• semantic blogging
• semantic wikis (structuring and browsing the web and desktop)
• social semantic collaborative filtering (using explicit relationships for information delivering and assessment)

A few short notes on each of these:

“Semantically-Interliinked Online Communities (SIOC)”

• motivation: there’s lots of latent information to be gleaned from all these socially enabled websites, but this information is hidden (there’s the underlying database, but all we see is the HTML. We can write wrappers, but when HTML changes we’re sore outta luck).
• SIOC is trying to expose the underlying structure via RDF, via plugins
• this stuff is getting integrated into lots of open source projects via plugins (phpBB, wordpress, drupal, more)

Semantic Blogging

• instead of just blogging for human eyes, blog for machine eyes too. e.g. automatically tag ppl with foaf info, or events with event xml, so that ppl can automatically add events to calendar or ppl to address book
• While this might be useful, I’m not sure if it will be the panacaea that its proponents claim. And, how are you going to get everyone to agree on standards? Are Microsoft, Google, and all the rest of the biggies going to agree to cooperate?

Semantic Wikis

• Addresses traditional wiki problems of structured access and information re-use
• of all the things he’s talking about, i think this one has the most potential for immediate adoption–in fact, it’s already being adopted in the small picture: categories and templates on wikipedia.
• However, I’m not sold on it being implemented on the grander scale. It looks like a lot of work, both for the readers and writers of the information (tools will make this easier, granted), but I also don’t think it will have the reader-base that eyeball wikis have. And I suspect it will be harder to motivate people to contribute information (more detached/less “social”)

Perhaps I’m being overly harsh. I WANT this to work, though. I’d love to see all this happening. But, I’m afraid it’s all a new re-release of the old “Semantic Web 1.0″ hype, with a shiny new rounded-corners, pastel-colored-gradient logo (and a “beta” on the side to boot). We’ll see…

If we are going to get a successful “Semantic Web 2.0″, I think we need to take a few lessons from the successes of the social software that has succeeded (these unfortunately were not described in the talk):

• Be Bottom-Up: Like folksonomy, don’t rely on the Powers that Be (W3C, Microsoft, Google, whoever) to set standards. Instead, let consensus bubble-up, folksonomy style. Users are lazy, so make doing the Right Thing easier than the Divergent Thing (see del.icio.us’s auto-suggest of tag labels for a good example of hwo this works). The big thing I’ve been wondering lately is if we can have a folksonomy of standards as well as a folksonomy of data–Will this work??
• Be Top-Down: Find a use that everyone really needs/wants, and they’ll jump through the hoops. Drawing from the Folksonomy example again, look at what Del and Flickr have done… Users are all functioning selfishly–they want to store their pictures and their bookmarks, and when they tag things, a lot of the time they’re only doing it for personal use. The folksonomic patterns bubble up on their own.

• Ultragleeper, by the guy who made Beautiful Soup. A “Recommendation Engine”. Ultragleeper takes group data from technorati, del, google, and personal recommendations regarding readworthiness of web pages. This is a neat subset of what I’m looking at for a good feedreader. I need to browse his source (python, yay) to take a look at how he implements ranking algorithms. I like that he bootstraps recommendations from the user’s OMPL feed list and delicious links… what a neat idea!
• Statistical analysis of “Front Page” Digg entries aggregated over time shows that while a small percent of users are responsible for a high percentage of the posts that make it to the front page, these prolific users are only responsible for posts that are in the mediocre-to-good range (they easily break the 50+ diggs requirement to hit the front page, but usually end up with a final rating in the < 500 range by the time they scroll off the page). The real high-scoring entries are typically submitted by relative unknowns. I need to think more about how this relates to personal recommendations.
• (off that last note… I need to consider if group recommendation sites (digg, delicious’s front page) have different demands/qualities/characteristics compared to personal recommendation requirements. Wider vs denser range of interests (a homogenous group would have much denser interest fields than an individual–this is why the delicious popular page is filled with ajax and css links), different tolerance of signal vs noise (is this true?), …