WHY OWL?
The current state of information technology in the modern enterprise has been described as a “Software Wasteland”.1 There are countless silos where each application has its own database and its own database schema with consequent duplication and high costs of integration and change. There are a few root causes.
First, there is no mechanism for breaking up a data schema into modules that can be re-purposed and reused across multiple databases. Monolithic data models and the lack of reusability increase the cost of change. Second, there is no way to uniquely identify data or schema elements globally across databases; this results in high integration costs. Finally, and perhaps most importantly, there are no widely used technologies and practices for representing the meaning of the data and schema elements as they evolve. Conceptual models may exist, but they are rarely kept up to date.
After a slow incubation period of nearly 15 years, the modern enterprise is waking up to the value of the Semantic Web stack of technologies, which has addressed all 3 of the above root causes.
The meaning of the data in a semantic application is defined using the Web Ontology Language (OWL). An OWL ontology is a model that represents the subject matter of the data (e.g., healthcare or electrical products) that will reside in triple-store databases that will be used by one or more related applications. OWL is built on the Resource Description Framework (RDF) which is a knowledge graph language based on triples. The support for globally unique identifiers is baked into RDF and thus OWL.
OWL is an essential ingredient in the semantic technology stack that continues to grow and evolve.
Leading-edge enterprises are building their own enterprise ontologies and enterprise knowledge graphs.2,3 The technology stack includes a graph query language analogous to SQL called SPARQL for querying triple stores. More recent additions to the stack include R2RML for converting data from relational databases into triples and SHACL for representing constraints and other details that are pertinent to specific applications. These are separate important tools for use in conjunction with OWL ontologies, and are beyond the scope of this book.
Finally, there are ample industry-scale tools provided by vendors that support these standards. The time is now to step out of the software wasteland.
WHY THIS BOOK?
It’s not easy to learn OWL on your own, even if you have a Ph.D. in artificial intelligence and training in formal logic. I found that out the hard way in 2002, when I was tasked with learning a precursor to OWL at a research lab at The Boeing Company. The purpose of this book is to dramatically speed up that learning process for others.
In 2010, I joined Semantic Arts as a senior ontology consultant and have been teaching OWL and using it to build industrial ontologies for the past seven-plus years. As few as five years ago, one of our clients told us not to use the “O” word (ontology)—because it would scare people. Back then, hardly anyone in a typical enterprise knew much about ontology, and there were few if any projects going on. That has dramatically changed in the past five years.
The need for ontologists is growing faster than the number and variety of available resources for learning OWL, especially from an industry perspective. This book differs from others available at the time of writing in that it is focused primarily on the needs of the modeler in an industrial context. I take a demand-pull approach, only introducing an OWL construct when the need for it arises to meet a modeling need. I focus on the 30% of OWL that gets used 90% of the time. Finally, I use examples from real-world industrial ontologies created in my day-to-day work.
The material in the first two chapters of this book has been presented five times in the past four years as a half-day tutorial. Venues included the Semantic Technology Conference, International Semantic Web Conference, Semantic Technology for Intelligence, Defense and Security, Data Architecture Summit, and Enterprise Data World. Some of the material has been folded into the Designing and Building Business Ontology class that I co-teach for Semantic Arts. The success I had with this material inspired me to expand it into a book.
TARGET AUDIENCE
This book is a gentle but thorough introduction to the most important parts of OWL. The only prerequisites for this book are an interest in modeling and a knack for thinking logically. The primary audience consists of modelers who want to build OWL ontologies for practical use in enterprise and government settings. For them, I drive most of the learning from real-world examples and avoid unnecessarily technical language. Secondary target audiences include the following.
• Intermediate to expert modelers in any setting having some familiarity with OWL who wish to deepen their understanding and see things from a fresh perspective.
• Technically oriented managers who want to know about ontology and OWL to better interact with their technical people.
• Undergraduates and post-graduates who want to understand OWL from a practical enterprise perspective.
• Instructors who are looking for new ways to explain OWL.
• Semantic technology developers who want a better understanding of the OWL ontologies that they code to.
OVERVIEW OF THIS BOOK
The book unfolds in a spiral manner. In the first cycle, I describe the core ideas. Each subsequent cycle reinforces and expands on what has been learned in prior cycles and introduces new related ideas. The book is divided into three parts.
Part 1: Introducing OWL
This is a cook’s tour of ontology and OWL, giving an informal overview of what things need to be said to build an ontology, followed by a detailed look at how to say them in OWL. This is illustrated using a healthcare example. I conclude by explaining some foundational ideas about meaning and semantics to prepare for going deeper in the next section.
Part 2: Going into Depth: Properties, Classes, and Inference
Everything to do with building an OWL ontology revolves around properties and classes. I give detailed descriptions of the main constructs that you are likely to need in everyday modeling, including what inferences are sanctioned.
Part 3: Using OWL in Practice
Using examples in healthcare, collateral, and financial transactions, we put into practice what we have learned so far. For each, I describe a model and show some inferences. Next, I identify some key limitations of OWL and possible workarounds. I conclude with a variety of practical tips and guidelines to send you on your way.
Per common practice, most of this book is written using the editorial “we.” At times, “we” will refer to the collective shared views and experiences of myself and my ontologist colleagues. “I” is used to express a personal perspective that may not be shared by my colleagues.
EXERCISES
In a number of places throughout the book you will find exercises. Answers may be found in Appendix A.4.
1 Software Wasteland by D. McComb, https://technicspub.com/software_wasteland/,
2 Exploiting Linked Data and Knowledge Graphs in Large Organisations http://www.springer.com/us/book/9783319456522.
3 Linking Enterprise Data. http://www.springer.com/us/book/9781441976642.