Friday, November 26, 2004

Dangling Databases

Why Relational Databases And Semantics Don't Mix "Jarg Corporation, which takes its name from "jargon", is about the next evolution of search. Actually, it is about more than that since semantics has wider applicability than search, but we will stick with search as an easily understood example of this technology."

"So, the question is: how does it do that?

The first answer is that it doesn't - at least in any general sense - only where it has already built an ontology which, in this case, is within healthcare. Indeed, its first customer is a hospital medical library. However, industry knowledge bases are becoming widely available and Jarg reckons that about 75% of the work involved in creating an ontology can be automated, so extending its product for new customers should not be a big issue."

"The second answer is that it achieves this sort of performance by refusing to use a relational database. Instead, it has patented its own approach, which involves storing semantic fragments. A segment fragment is either two elements and the relationship that joins them (for example, "Waterloo is a station") or it can store and element with a "dangling" relationship. This latter concept is especially important. The whole point about searching is that you want to be able to discover relationships that you didn't know existed."

No comments: