Earlier this week, I pointed out a Bio-IT World article relating how Pfizer was using semantic technologies to circumvent data silos for better pharmaceutical research.

They'd built a system called PEKE, and as I shared in the blog post, the man who headed the project, Ted Slater, realized he couldn't integrate data silos as quickly as business users could create them. So, PEKE was designed to make data accessible wherever it resided, without actually providing point-to-point integration.

The article gave a birds-eye view of the technology at work, but even so, I suspect the very term “semantic integration” might have made a few readers' eyes glaze over.

As luck would have it, right after that post, I happened to have an interview with David Linthicum. He is now well-known as a SOA expert and consultant, but he also has an extensive background in enterprise application and data integration. I mentioned the article to him, pointing out that I didn't really understand the difference between interoperability and integration.

His response: In short, there is no difference. Whenever he's looked into interoperability, he found there's always integration underlying it.

This response surprised and frustrated me – although, given how often terms and definitions change in this business, I'm not sure why I was surprised. As if to further illustrate the ... er ... semantic games the industry plays, Linthicum pointed out that what some people refer to as semantic understanding is also called a common data model.

I suggested it'd be a great topic for him to explain in a blog post, and guess what? On Thursday, ebizQ published, “The Place for Semantics within Data Integration,” written by Linthicum. Now that's service! (Cable companies, are you listening?!?)

The piece takes a lot of the mystery out of semantics and how it can be applied to integration. He points out, first of all, that semantics is not integration or a replacement for it, but rather a tool you can use on your way to integration:

“So, what semantic tech can do is to create a common semantic understanding of an integration domain, but typically does not provide the mechanisms for integration unto itself, including extraction from the source, transformation, translation, routing, transport, and production to the target.”

Vendors are starting to toss the term semantic technology around a bit, and I suspect it will only get worse. ebizQ reader Chuck Allen has noticed this trend too, and wrote this response to Linthicum's post:

“Consultants tend to oversell whatever they are selling and lately I've heard a few pitches that seem to assert that the right ontology or semantic technology is the cure for whatever might be the customer's integration pain. There is a lot to be said to having a solid semantic understanding and agreement regarding data, processes, etc., but as your post points out, there's more to integration than a good semantic foundation.”

Good to know. I hope this information will help those of you in the field avoid the fast-talk of semantic hype.