Killer Apps for the Semantic Web

A "killer app" (application) is the one tool or usage that propels a platform forward. It is the usage that compels the early adopters to use a platform. In the early days of personal computers (PCs), killer apps were actually applications. Since then the term has become generic, meaning any compelling reason to adopt a new platform technology. Once a platform is established, more mainstream users are brought in (largely through word of mouth from the early adopters), and many other uses are found. Visicalc was the killer app for the early PCs in that many of the early adopters bought PCs just to run the spreadsheet. File sharing was the killer app that brought local area networks (LANs) and Novel into the mainstream. Universal (not just within your own company) email was the killer app for the Internet, and academic research (hypertext and search) was the killer app for the Web.

I believe that there are two killer apps for the Semantic Web: better query and agents. (Disclaimer: Killer apps are always easy to spot retrospectively, but they are very iffy prospectively.)

Better Query

The big payoff for the Semantic Web will be the ability to execute a conceptual query over the entire Web.

There are two kinds of queries in the world now: structured and unstructured. Structured query language (SQL) and Google are, respectively, the best known exemplars of each. SQL yields precise results for small domains in individual databases if you know the schema; Google yields broad and expansive results over the domain of reachable indexable text on the Internet.

However, neither is adequate, either now or (more important) in the future. They both rely too much on humans. The structured query relies on the human who is doing the requesting knowing three things: where to go (which database), what to ask for (in the vocabulary of the database schema), and how to ask (the query syntax). Although some of this can be hidden with user interfaces, most of these problems are still present.

The unstructured queries require us to be clever about what we ask for and to sift through what we get. While searching for a paper by Uche Ogbuji, I was pleasantly surprised to discover that he has what appears to be a globally unique name—all 257 entries that come up on a search refer to the same person (who, by the way, has written many excellent articles about XML and the Semantic Web). However, a search for something by Roger Smith requires far more sifting on the back end.

Try finding an article written by Hillary Clinton instead of one written about her. And we haven't even gotten to complex queries (e.g., find homes within a quarter mile of a fault line, or find a used car that gets 35 miles per gallon). There just isn't enough cleverness to construct keywords that will have the right amount of proximity and not exclude many good entries. Try finding something that the author of a Web page didn't know, or using terms the author didn't use (that's what our example later in the chapter will cover).

Agents

The Web is a big place, and it's getting bigger all the time. We are incredibly fortunate to have Google, which has managed to keep a centralized copy of all the statically reachable parts of the Web. We have to keep in mind, though, that Google has 10,000 servers dedicated to this task. We also have to keep in mind that a huge percentage of the potentially available information is in databases that are attached to the Web, but are not indexable by Google (or anyone else). You must go to those sites with your specific query, expressed in their language, or fill in their form (supplying your ID and password) to see if they have anything that matches your request.

Agent

An agent is a program to which an individual delegates some authority to act on the individual's behalf and the releases to act autonomously.

I believe we will see a lot more of agents in the near future. In the early days of the Internet we were much more aware of them; gopher, archie, and veronica were often used as agents to find things on the Internet, before there were comprehensive indexes. But even now agents (spider, bots, and the like) are crawling the Web as we speak, updating indexes and looking for stuff.

The Semantic Web promises to make possible end-user, task-specific agents. At one level, agents could be long-duration distributed queries. They could also transact on your behalf.

We'll discuss the challenges and the "enables" to getting agents into the hands of everyday users in the section on adoption.