In 2003, when the World Wide Web Consortium was working toward the ratification of the Recommendations for the Semantic Web languages RDF, RDFS, and OWL, we realized that there was a need for an industrial-level introductory course in these technologies. The standards were technically sound, but, as is typically the case with standards documents, they were written with technical completeness in mind rather than education. We realized that for this technology to take off, people other than mathematicians and logicians would have to learn the basics of semantic modeling.
Toward that end, we started a collaboration to create a series of trainings aimed not at university students or technologists but at Web developers who were practitioners in some other field. In short, we needed to get the Semantic Web out of the hands of the logicians and Web technologists, whose job had been to build a consistent and robust infrastructure, and into the hands of the practitioners who were to build the Semantic Web. The Web didn’t grow to the size it is today through the efforts of only HTML designers, nor would the Semantic Web grow as a result of only logicians’ efforts.
After a year or so of offering training to a variety of audiences, we delivered a training course at the National Agriculture Library of the U.S. Department of Agriculture. Present for this training were a wide variety of practitioners in many fields, including health care, finance, engineering, national intelligence, and enterprise architecture. The unique synergy of these varied practitioners resulted in a dynamic four days of investigation into the power and subtlety of semantic modeling. Although the practitioners in the room were innovative and intelligent, we found that even for these early adopters, some of the new ways of thinking required for modeling in a World Wide Web context were too subtle to master after just a one-week course. One participant had registered for the course multiple times, insisting that something else “clicked” each time she went through the exercises.
This is when we realized that although the course was doing a good job of disseminating the information and skills for the Semantic Web, another, more archival resource was needed. We had to create something that students could work with on their own and could consult when they had questions. This was the point at which the idea of a book on modeling in the Semantic Web was conceived. We realized that the readership needed to include a wide variety of people from a number of fields, not just programmers or Web application developers but all the people from different fields who were struggling to understand how to use the new Web languages.
It was tempting at first to design this book to be the definitive statement on the Semantic Web vision, or “everything you ever wanted to know about OWL,” including comparisons to program modeling languages such as UML, knowledge modeling languages, theories of inferencing and logic, details of the Web infrastructure (URIs and URLs), and the exact current status of all the developing standards (including SPARQL, GRDDL, RDFa, and the new OWL 1.1 effort). We realized, however, that not only would such a book be a superhuman undertaking, but it would also fail to serve our primary purpose of putting the tools of the Semantic Web into the hands of a generation of intelligent practitioners who could build real applications. For this reason, we concentrated on a particular essential skill for constructing the Semantic Web: building useful and reusable models in the World Wide Web setting.
Many of these patterns entail several variants, each embodying a different philosophy or approach to modeling. For advanced cases such as these, we realized that we couldn’t hope to provide a single, definitive answer to how these things should be modeled. So instead, our goal is to educate domain practitioners so that they can read and understand design patterns of this sort and have the intellectual tools to make considered decisions about which ones to use and how to adapt them. We wanted to focus on those trying to use RDF, RDFS, and OWL to accomplish specific tasks and model their own data and domains, rather than write a generic book on ontology development. Thus, we have focused on the “working ontologist” who was trying to create a domain model on the Semantic Web.
The design patterns we use in this book tend to be much simpler. Often a pattern consists of only a single statement but one that is especially helpful when used in a particular context. The value of the pattern isn’t so much in the complexity of its realization but in the awareness of the sort of situation in which it can be used.
This “make it useful” philosophy also motivated the choice of the examples we use to illustrate these patterns in this book. There are a number of competing criteria for good example domains in a book of this sort. The examples must be understandable to a wide variety of audiences, fairly compelling, yet complex enough to reflect real modeling situations. The actual examples we have encountered in our customer modeling situations satisfy the last condition but either are too specialized—for example, modeling complex molecular biological data; or, in some cases, they are too business-sensitive—for example, modeling particular investment policies—to publish for a general audience.
We also had to struggle with a tension between the coherence of the examples. We had to decide between using the same example throughout the book versus having stylistic variation and different examples, both so the prose didn’t get too heavy with one topic, but also so the book didn’t become one about how to model—for example, the life and works of William Shakespeare for the Semantic Web.
We addressed these competing constraints by introducing a fairly small number of example domains: William Shakespeare is used to illustrate some of the most basic capabilities of the Semantic Web. The tabular information about products and the manufacturing locations was inspired by the sample data provided with a popular database management package. Other examples come from domains we’ve worked with in the past or where there had been particular interest among our students. We hope the examples based on the roles of people in a workplace will be familiar to just about anyone who has worked in an office with more than one person, and that they highlight the capabilities of Semantic Web modeling when it comes to the different ways entities can be related to one another.
Some of the more involved examples are based on actual modeling challenges from fairly involved customer applications. For example, the ice cream example in Chapter 7 is based, believe it or not, on a workflow analysis example from a NASA application. The questionnaire is based on a number of customer examples for controlled data gathering, including sensitive intelligence gathering for a military application. In these cases, the domain has been changed to make the examples more entertaining and accessible to a general audience.
We have included a number of extended examples of Semantic Web modeling “in the wild,” where we have found publicly available and accessible modeling projects for which there is no need to sanitize the models. These examples can include any number of anomalies or idiosyncrasies, which would be confusing as an introduction to modeling but as illustrations give a better picture about how these systems are being used on the World Wide Web. In accordance with the tenet that this book does not include everything we know about the Semantic Web, these examples are limited to the modeling issues that arise around the problem of distributing structured knowledge over the Web. Thus, the treatment focuses on how information is modeled for reuse and robustness in a distributed environment.
By combining these different example sources, we hope we have struck a happy balance among all the competing constraints and managed to include a fairly entertaining but comprehensive set of examples that can guide the reader through the various capabilities of the Semantic Web modeling languages.
This book provides many technical terms that we introduce in a somewhat informal way. Although there have been many volumes written that debate the formal meaning of words like inference, representation, and even meaning, we have chosen to stick to a relatively informal and operational use of the terms. We feel this is more appropriate to the needs of the ontology designer or application developer for whom this book was written. We apologize to those philosophers and formalists who may be offended by our casual use of such important concepts.
We often find that when people hear we are writing a new Semantic Web modeling book, their first question is, “Will it have examples?” For this book, the answer is an emphatic “Yes!” Even with a wide variety of examples, however, it is easy to keep thinking “inside the box” and to focus too heavily on the details of the examples themselves. We hope you will use the examples as they were intended: for illustration and education. But you should also consider how the examples could be changed, adapted, or retargeted to model something in your personal domain. In the Semantic Web, Anyone can say Anything about Any topic. Explore the freedom.
Second Printing: Since the first printing there have been advances in several of the technologies we discuss such as SPARQL, OWL 2.0, and SKOS that go beyond the state of affairs at the time of first printing. We have created a web site that covers developing technology standards and changing thinking about best practices for the Semantic Web. You can find it at http://www.workingontologist.org/.
ACKNOWLEDGMENTS
Of course, no book gets written without a lot of input and influence from others. We would like to thank a number of professional colleagues, including Bijan Parsia and Jennifer Golbeck, and the students of the University of Maryland MINDSWAP project, who discussed many of the ideas in this book with us. We thank Irene Polikoff, Ralph Hodgson, and Robert Coyne from TopQuadrant Inc., who were supportive of this writing effort, and our many colleagues in the Semantic Web community, including Tim Berners-Lee, whose vision motivated both of us, and Ora Lassila, Bernardo Cuenca-Grau, Xavier Lopez, Zhe Wu and Guus Schreiber, who gave us feedback on what became the choice of features for RDF-PLUS. We are also grateful to the many colleagues who’ve helped us as we’ve learned and taught about Semantic Web technologies.
We would also especially like to thank the reviewers who helped us improve the material in the book: John Bresnick, Ted Slater, and Susie Stephens all gave us many helpful comments on the material, and Mike Uschold of Boeing made a heroic effort in reviewing every chapter, sometimes more than once, and worked hard to help us make this book the best it could be. We didn’t take all of his suggestions, but those we did have greatly improved the quality of the material, and we thank him profusely for his time and efforts.
We also want to thank Denise Penrose, who talked us into publishing with Elsevier and whose personal oversight helped make sure the book actually got done on time. We also thank Mary James, Diane Cerra, and Marilyn Rash, who helped in the book’s editing and production. We couldn’t have done it without the help of all these people.
We also thank you, our readers. We’ve enjoyed writing this book, and we hope you’ll find it not only very readable but also very useful in your World Wide Web endeavors. We wish you all the best of luck.