Nov. 2013 (Vol. 46, No. 11) pp. 8-9
0018-9162/13/$31.00 © 2013 IEEE
Published by the IEEE Computer Society
Published by the IEEE Computer Society
Ian Horrocks: Standardizing OWL
|Growing Up and Out|
PDFs Require Adobe Acrobat
Ian Horrocks describes the early days of the Web Ontology Language (OWL) and the effort it took to standardize it and other languages.
Early research into artificial intelligence essentially boiled down to capturing “knowledge” and making it available to software in a format that would allow that software to behave more intelligently. For quite a long time, the syntax and format used to record or enter this knowledge into files was tied to the very programs that would read and use it. Research groups would typically develop tools and define a format to feed knowledge into them. The hope was to build reusable reasoning tools that could function across many domains of knowledge by simply loading different knowledge sets into those systems.
In the late 1990s and early 2000s, AI researchers realized that to maximize the usefulness of their work in an increasingly networked environment, they needed to standardize their ontology/knowledge representation languages and move the focus from “yet another syntax to represent knowledge” to software that could evolve and use the knowledge to exchange data between different applications.
I recently interviewed Oxford University's Ian Horrocks about how the various ontology efforts in the late 1990s were brought together, standardized, and normalized to produce the Web Ontology Language (OWL). You can view our full conversation at www.computer.org/computingconversations.
Initially, those working in the AI field came from a very narrow area that was simply trying to capture and codify knowledge:
My background had been in medical informatics and developing what we now call “ontology languages” and reasoning systems, although we weren't necessarily calling them ontologies back then.
The first step toward developing a standard ontology language was a series of small meetings in which people simply shared their best ideas and learned about other approaches to knowledge representation:
In 1999, I went to an ontology-sharing day in Kaiserslautern and met people like Frank van Harmelen and Dieter Fensel who were also working in this area. I managed to convince them that description logic would be a good starting point. It's a type of logic whose rationale is to formalize what we now call ontology languages. Compared to frame languages, description logic had more expressive power and a very clear formal semantics because it was basically a fragment of first-order logic.
The small group developed a description logic-based syntax and language that they called OIL (Ontology Interchange Language) and published their approach (“OIL: An Ontology Infrastructure for the Semantic Web,” IEEE Intelligent Systems, March/April 2001, pp. 38-45). As these small collaborative efforts were more broadly shared and published, more researchers became interested in the shared objective of growing the overall field through the use of standard ways to represent knowledge:
We met people in the US like Peter-Patel-Schneider, Pat Hayes, Jim Hendler, and others working on the DAML [DARPA Agent Markup Language] program. We all decided that we were more or less trying to do the same thing, so why not pool our resources? Accordingly, we formed the rather grandly named “Joint US/EU ad hoc Agent Markup Language Committee” and produced DAML+OIL. It wasn't really much different from the OIL specification. The idea was to draw more stakeholders into the process so we could develop DAML+OIL into a standard. This was where the OWL effort and working group started.
Growing Up and Out
With EU and US researchers finding that their approaches and goals were well aligned, they set about working with the W3C (World Wide Web Consortium) to produce a formal standard based on the DAML+OIL approach. They initially expected that with DAML+OIL well developed, it would be rather simple to wrap up the standardization effort in a short amount of time.
But as the group grew, more people learned about the work, became interested in the resulting specification, and wanted to participate:
A whole new bunch of people joined the party, in particular people from the Web community such as Dan Connolly and Sandro Hawke. They had a whole load of concerns of their own and things that were important to them, such as integration and compatibility with RDF [Resource Description Framework], general Web infrastructure, and existing standards.
The Web Ontology (WebOnt) Working Group within the W3C formed in November 2001, just two years after the ontology-sharing day back in 1999. But in that brief period, many new people had come to the table, and these different stakeholders had needs to be addressed in the resulting specification:
The process of evolving DAML+OIL into OWL took longer than we thought and involved a bigger change than we thought. It took a couple of years in the end—and probably 10 years off my life.It was interesting, and I learned a lot as the language evolved. It was mainly the syntax and the relationship with RDF that changed. The underlying logic didn't change very much, nor did the semantics of OWL DL because it just flows from the logic.
Even though the WebOnt Working Group needed to address the needs of myriad stakeholders, by December 2003, the standard was complete and published:
There were huge arguments, but in the end, we managed to reach a compromise that everybody could sign off on, some with more grumbling than others.
With a standard language to represent knowledge, the field started to come together and produce tools and data that could be shared across research projects and commercial efforts:
We had a standard KR [knowledge representation] language that was supported by lots of different groups building infrastructure, and suddenly applications people started to feel more comfortable. It had always been an issue that you were using a system from the University of X and then the relevant research project ended or the research group got bored and went off and did something else, leaving you with no support. But with OWL, you could use a standard language, with tons of people developing an ever-growing array of infrastructure to support building, deploying, and maintaining ontologies.
As the field went from building one-off toolsets tied to specialized KR formats to an increasingly solid and reusable infrastructure, a wide range of researchers in computer science and other scientific fields could focus on solving the more interesting problems. Today, OWL and RDF are integrated into so many toolsets that users barely notice them:
What amazes me today with both OWL and RDF is that you bump into people all the time who are building applications with those tools and infrastructure that you don't know. The stuff just works now, so they can download it off the Web and build it into applications. Users are pretty happy.
The efforts to standardize RDF and OWL have produced unintended benefits for our field. Those who use these technologies daily typically have no idea about the innovators who took the time to reach across research projects to come up with a unified and general-purpose approach that was widely usable:
As a community, we haven't been very good about embracing it as a great success and saying, “We've done a really good job here—we built stuff that people are using and it works.” It works off the shelf these days—the tools are pretty robust, and we should be proud of that as an achievement of the community.
For more information about OWL, please see www.w3.org/2001/sw/wiki/OWL.
Charles Severance, Computing Conversations column editor and Computer's multimedia editor, is a clinical associate professor and teaches in the School of Information at the University of Michigan. Follow him on Twitter @drchuck or contact him at firstname.lastname@example.org.