One of the afternoon sessions at the Web 2.0 Summit I’m most looking forward to is the “Semantic Web” panel, in part because it’s a hot topic and in part because it’s the lead topic in the most recent Release 2.0. We’ve long been interested in the diverse work of Danny Hillis, one of the people behind Metaweb and its Semantic Web project Freebase. We spoke recently to him. Here are a couple of highlights from our conversation:
How has Freebase changed your ideas about the Semantic Web?
The personal surprise has been how much energy is coming from commerce, not science. Everyone else — that is, everyone who isn’t a scientist like me — would have known that. Data sharing is what scientists do for a living. Scientists tend to have narrow communities with which they share data. The data sharing problem for them is easier than it is for the general public. Film information, restaurant information: all that is shared with a much wider community than, say, proteomics information. That’s the shift for me. It presents a bigger, more interesting problem.
How do you know when something needs to be organized and something needs to be free-form?
It’s not a binary question. It’s more about where along the spectrum you want to be for different kinds of data. And there are issues related to those tradeoffs. How quickly can you get started? The Semantic Web offers a powerful, general-purpose set of tools. There are so many things you can do, so many ways you can organize. There are an embarrassment of options. Databases in general require an awful lot of thought up front. A committee decides the schema, and it’s inflexible after that. Most relational databases you design up front and live with the consequences. You have to devote energy to getting around the decisions you made up front. Free text dominated on the Internet. There was no database, just text. The center of what’s happening on the web is something between the two. Until recently, the only pattern was hyperlinks. Google mined that and accomplished a fantastic amount with that. Google showed how much help a little structure could bring to navigating things. Now Metaweb and others are seeing ways to bring in more structure, but not as much as a relational database where we have to agree before we get started. Tagging is one step away from a total non-structure. It’s an intermediate step.
We’ll hear more from Hillis and his colleagues/competitors this afternoon.