What's Holding Up Semantic Integration?

Loraine Lawson

Last year, Gartner predicted that semantic technology would emerge as one of the top 10 disruptive technologies within the next four years.


That's right: I'm keeping tabs on that prediction.


I've written on several occasions about how semantic technology can be used in integration, a field known as semantic integration. It's a fascinating emerging technology to follow, but so far, it's still outside the mainstream, used primarily by large tech companies, pharmaceutical companies and other large, cutting-edge organizations, including NASA.


Still. We can dream, can't we?


Recently, Dr. Dobb's Journal interviewed Richard Keller, senior research computer scientist and group lead for the information sharing and integration group at NASA, about semantic integration. It's a fascinating discussion, and I particularly liked that Keller addressed one critical questions I've long wondered about: What's it going to take to make semantic integration possible for more companies?


Granted, that isn't the question he was asked, so don't try skimming the interview to find it. The question he was asked was, "What's the next big hurdle to achieving semantic integration?" He gave a two-part answer, responding that ontology mapping is the first big hurdle, and it's long been an issue for the semantic Web community.


It's the second part of his answer that speaks to my long-standing question:

More broadly, I think the challenge for making semantic integration work in the marketplace is to make it quicker and easier to specify data semantics. Currently, specifying semantics using ontologies is a somewhat arcane and tedious process. ...Although we are starting to see some good tools on the market to make this process easier, the cost/benefit calculations are not yet sufficiently favorable to support widespread adoption of this approach.

Keller explains the challenges and opportunities with semantic integration in a concise, accessible manner. He explains why standards aren't the answer to ongoing data-integration problems-they're onerous and expensive-and how the W3C's SPARQL will help but not solve the difficulty of semantic integration. The W3C's SPARQL applies to querying, but as Keller explains, that's not the underlying difficulty. Construction of the ontologies is tricky with semantic integration. That's why the maps are so critical, and so difficult.


Definitely check out this short interview if you're interested in semantic integration. You might also find the following IT Business Edge posts helpful in understanding semantic technology and its impact on integration:
Semantic Integration Is Emerging Field - Should You Care?
Myths About Semantic Technology Part I and Part II
Can Semantics Tech Eliminate the Need for Data Integration?

Add Comment      Leave a comment on this blog post
Oct 27, 2009 2:00 AM Stijn Christiaens Stijn Christiaens  says: in response to JP Morgenthal

Correct - semantics is about getting agreements between humans (i.e., do we mean the same thing when we use the term "customer"). This, together with the issue that easy, mainstream tooling is not yet out there (and less focused on the agreement side), is part of the reason that the adoption rate is slow ... but nonetheless steady based on the market feedback we get at Collibra.

Next to tools, it is also important to provide the right methods (e.g., http://deleenheer.wordpress.com/business-semantics-management/ ), where a mix of roles and responsibilities guarantee that the organization's business semantics come into existence - and stay that way.

About the prediction: I'm keeping tabs as well

Oct 27, 2009 11:51 AM pedanticweb pedanticweb  says:

Loraine-- I think Keller's point is largely a cop-out.  Take for instance your typical BI project based on conventional technology; the specification of ETL mappings, dimensional models, aggregate functions, and multi-way join SQL to correctly populate dashboards are no less "arcane and tedieous" than working with ontologies.  The part where he gets it right is about tooling being immature. If Semantic Web tools had one-tenth the investment made in conventional data integration technology they would already be disruptive. The more interesting question is why haven't major data integration tools vendors invested more in Semantic Web technology -- the answer to that question will certainly have more to do with economics than with the relative goodness of the technology.

Oct 27, 2009 12:10 PM JP Morgenthal JP Morgenthal  says:

I released "Enterprise Information Integration: A Pragmatic Approach" in 2005, which focuses heavily on the semantic integration problem domain.  Since then the book has sold steadily, but at a low volume.  I have taken that to mean that there is a slow swelling of need here, but it's having a heck of time rising up to a mainstream technology. 

Semantic Web does not require SPARQL to succeed, although, inference is a great way to leverage and access the information in a Semantic Web.  First and foremost, semantics is about meaning of words and it seems we get further and further away from agreement on words than closer together.  While the Semantic Web accounts for this disparity, the results are less than stellar.

Semantic integration is very possible.  I proved it back in 2004 with CompuCredit and SIRVA, but it requires agreement of humans.  I guess that's a human task, so we're probably just missing BPEL4People in the mix .

Oct 30, 2009 10:51 AM Michael Waclawiczek Michael Waclawiczek  says:

Interesting dialog.  We think that semantic data integration is a very important topic and a key enabler for any company that wants to deal with the crucial pain point in data integration -- how to simplify data mapping design, enable business rule reuse, and provide a common language for business and IT users involved in a data integration project.  We've taken a very practical approach to semantic data integration and our smart semantics approach is increasingly becoming a key differentiator for us.


Post a comment





(Maximum characters: 1200). You have 1200 characters left.



Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.