Dave McComb:(Mr. McComb is president of Semantic Arts and author ofSemantics in Business Systems.) Here are the three issues I picked. If you live and work in a corporation or large government entity, you live in what I call flat land. You've got this gigantic flat schema space, where you've got tens, hundreds of thousands of attributes with names like claimant's, claim, update, date. Just bizarre, long strings of highly abbreviated stuff, and when you bought packages or whether you did it yourself, you've got hundreds of thousands of these. In order to get any useful work done, you've got to know what they all mean.
We can sit here and think, "Oh surely it's not that bad." It's way worse. I call that flat earth. There's almost no abstraction at all, other than a few people's heads. How did it happen? Because what happens is people did the central design—logic designs and physical designs to enable the system. It's in SQL, and all that stuff is thrown away, which was the abstraction, because SQL just works against the physical data. If you're going to do maintenance, you got to do it at the flat-land level. Everything at the flat land, they just threw it all away.
The other thing is hearing this categorization stuff: people who every time they come across something new make a new category, and people who every time they come across something new they put it in a category that already exists. We're only a little bit out of the woods on this. You know, operating adherence gives us a little bit, especially business odds—you can get some abstraction there—but most objects on your programming have been on the mechanics side, not on the business side. Using web services to encapsulate some of that helps. And what we're here for today.
-- ADVERTISEMENT --
Deborah initiated this discussion about size. If I were to take it further, I've gone all the way from the large to the incredibly small. It's hard now for me to find an ontology that's too small to be useful. Genealogies: You can imagine what's in there. Brothers, sisters, aunts, uncles, all that kind of crap. Those are not very big ontologies. But if you think about it, even, they are too big. And you can commit to it and you know how to breed all of it, make changes. But the certain micro-ontology behind it, it's just, what's the definition of what a person is; what's the difference between male and female? Especially the really top: If you agree on these three things it gives them what's a parent --> what's a parent. If you agree on those three, you can bring in the other ontology to figure out the aunts, uncles, brothers, and sisters.
If you can't even agree on those three, forget about it.
This is interesting and intriguing work. When you get out of the linguist's world it's called semantic primes. There's a small number of base concepts that we absolutely have to have figured out, or we're just not going to make it. Everything else we've done is based on top of that. And you get things like up and down. Every culture has a concept of up and down, because we grew up in a world of gravity. Near-far, inside-outside; there's like 40 or 50 of them. We're not mapping 10,000 words in English and 10,000 words in German. It's just a real small number.
Where I want to go I think is a similarly small number—very small in the world of business systems. And if we just agree on those, we're just building inside of that.
Jeff Pollock:(Mr. Pollack is vice president of technology for Network Interference.) One of the questions raised was what is the difference between the consumer-oriented semantic web en masse and the enterprise semantic web, the ways of using it in there. One of the key things is that inside the enterprise you're dealing a lot more with semi-structured data and highly structured data in legacy applications. So, you've got lots of EDI-type messages floating around; you've got lots of existing databases and COBOL applications and things that have 5-, 10-, 20-year histories. We're focusing on ways to improve the semantic content associated with these semi-structured/structured sources.
The value proposition of semantic technologies is pretty simple. It's about having fewer humans in the loop, more automation, and adaptive information structures. All that ties back of course to cost savings. Fewer humans, more automation is pretty much true at any level, thinking about the semantic web.
I'll run through a couple core business-use cases that we see at Network Inference as far as applied uses of semantic technologies. I'll also talk about why we're so heavily committed to OWL.
Why semantic web services are getting a lot of attention right now is this notion of meaning embedded in the plumbing, if you will, making a metaphor here that web services are in fact dealing with plumbing issues of moving messages around an enterprise. Looking at how meaning can be used in conjunction with web services is actually important, because it enables dynamic discovery of those services, dynamic collaboration, and automatic service orchestration. If you think about the amount of code that has to go into web services adapters and orchestration and any excess that goes into managing those interfaces, the applied use of semantic technologies in this arena will in fact automate web services.
If you think of semantic web services as dealing with meaning inside the plumbing of routing between enterprise applications, business inferencing is really about single applications leveraging semantics to support more adaptive systems. This notion of machine visibility is not new, but what is new is the idea that you can do it in a standard way. With OWL and RDF as a global recommendation through the W3C, you can now begin to build adaptive information structures inside your line of business applications and support dynamic business rules, dynamic data. We see a large proportion of our customers at Network Inference doing this, specifically building applications in a model-driven environment by using OWL as a way to derive business rules on the fly.
[Some] existing vendors have platforms that basically derive a lot of their embedded data manipulation software in legacy environments, and you can't simply plug in new semantic technologies—adapters, if you will, or modules—and receive dramatic improvements upon their existing infrastructures. We believe it's very important from the middleware perspective to support semantic data integration at the core level with OWL- and RDF-based structures, because that's where you derive a lot of the advantages with adaptability and flexibility. Improvements are always achieved in applying those two technologies.
What we really got from XML and RDF are machine-interpretable markup, so that allowed us to build parsers in programs that could deal with information that was going across the wire and take it apart and put it back together in new environments, new formats. Looking at DAMA and OIL, we're really talking about providing a machine-actionable semantic. These two sets of languages provide a semantic syntax, if you will. It's a way of defining relationships between objects at the data level. It's very precise, very flexible.
In the future there's going to be this notion of semantic web services where OWL will support at some level service descriptions, rules, formal logic rules, and possibly mappings—how we get all of these ontologies that are floating around in the world to relate to one another. Description logics might be a very nice way of doing that because it's all about building axioms between concepts.
I want to quickly talk about some of the other standards groups that are using OWL in core technologies. I happen to sit on the ebXML Semantic Content Management Group, where we're looking at specifying the core ebXML rim or the repository and the registry in the OWL interface. The reason for that is when you point reasoners and inference engines at repositories you could begin to infer new facts on the fly and achieve new dynamic, flexible results out of existing data sources that you couldn't do previously with highly structured query environments.
OWL is loosely coupled. We're able to take the semantics of data sets and decouple them from the actual instances and then work with them in a registry repository-type environment and still have the reference ability to tie back to what the enterprise instance data is.
If you can imagine an existing large corporation with hundreds of databases and applications and customer components, the OWL models can be a decoupled representation of those semantics and still point back to the original sources and provide a lot of benefits of tying back to that and the flexibility of the inferencing. They are easily federated, which is important, so this loose coupling aspect is an important thing in the IT ecosystem.
Tell us what you think of this post using our On or Off rating system. Only your most recent vote will count.
Member Comments
Small World,
Thank you very much for posting this article... just finished chatting with Dave H before logging in.
We Are Drowning in Information but Thursting for Knowledge.
John Naisbitt
Given the SQL world of thinking and licensing how can we best introduce this to the corporate world ?
How can we change thier behaviour and way of working ?
I have tried and it is not easy, but it makes so much sense to me as well.
Is there really a quest for knowledge with an alligenment to business processes, versus "uintelligent" flat structures, and at what cost ? ( time and effort)
Try talking to an SQL DBA about OOP. This is already challenging. Now try implementation and licensing politics.
I agree to build semantic models based on smaller applications and then to grow the model step by step, building your own ontology. technically I see no problems, XML, OO, mapping etc..
Jeff,Dave,Deborah.. I would like to talk to you, about how you implement at the customer site. I will send you an email, and if you feel comfortable, please feel free to add a few points here as well. I feel it really gets down to educating the customers and gradually waiting for a change in behaviour ( evolution) to occur. I too am tired of hearing customers say.. we have a global licensing agreement with XYZ company... maybe some day it will get down to competitive advantage and companies will decide to change their way of working.
Maybe in 20-30 years we will be building Knowledge Industries & businesses ( like todays services and manufacturing), based on applications & ontologies. I have another business idea.. thanks..
AP WorldStream via COMTEX
What`s Missing in the Paying Bloggers to Blog Vision ?
E-NEWSLETTER SERIES