(This was posted two days ago, but revealed a problem that crashed certain IE browsers. I’ve tried adjustments to account for IE’s CSS problems, with much testing help from Talan and Andrew, and we’ll see how it goes.)

Erik has made some great changes to All Consuming – which put to rest all my reservations of a week ago. First (as both he and Jill have blogged) it now understands links that include “isbn=0123456789″ as references to books – expanding the book comments it can aggregate to include people linking to any site that uses such a formulation (such as Booksense, B&N, and All Consuming itself). Second, All Consuming now doesn’t just send people to Amazon to buy books – but also offers links to Booksense (as one can see on any book detail page ).

Bravo! Now, how do we get others to follow this example?

Also, of course, there’s that bigger issue we’ve been discussing: “How do we know we’re all talking about the same thing?” (the same book, the same news item, etc). As Erik alluded in a comment on Jill’s blog, when we ask this question we’re headed toward the territory of the Semantic Web efforts (W3C, SemanticWeb.org). Given this, here are some further questions: How does the Semantic Web intersect with issues of link politics? Can we agree on an ontology even if we have quite different values?

I think these are interesting issues. I’m not the Semantic Web detractor I’ve sometimes been made out as (and sometimes misquoted as), though I do think the hype has distorted the presentation of the research. Frankly, I think efforts like All Consuming are closer to being the Semantic Web than many projects that bear the name. Why? Because All Consuming is actually web-centric, whereas most SM projects are about hooking databases up to each other (with the web as a human-read medium largely extraneous or even uninvolved). And I think it was Michael Mateas (who’s collaborating with Andrew on Facade) who pointed out to me that, as in other areas of AI, statistical approaches to understanding the Web (e.g., Google) keep clobbering human attempts at classification (e.g., Yahoo). Not to say the Open Directory isn’t useful – or that SM technologies won’t be useful – on the human-read web. I think we’re likely to see interesting hybrids as the SM efforts gain traction: the next steps beyond the single-database-driven sites we have now. In the meantime, people like Erik will keep doing an impressive job without SM tech, using the information encoded in our links to a multiplicity (hopefully) of those database-driven sites.

This entry was posted on Thursday, November 21st, 2002 at 5:13 pm and is filed under old-hyperfiction-blog. You can follow any responses to this entry through the RSS 2.0 feed. Responses are currently closed, but you can trackback from your own site.