CHAPTER5Modeling Knowledge in theReal World‘‘Welcome to the real world’’—Morpheus, The MatrixMost of the discussion so far about RDF, OWL, and ontologies has been veryabstract. To this point, you’ve learned what they are, but not as much abouthow to use them in practical applications and how to utilize the informationthey enable you to model.This chapter is about using OWL ontologies in the real world. In it, youlearn about:How ontologies fit into practical applicationsThe concept of inference and how it is critical to the implementationof the semantics of OWLProfiles of OWL, their purposes, and how they can be used to providedesirable computational characteristics to systems using OWLCritical design principles for information-management applicationsExploring the Components of the Semantic WebOWL ontologies allow you to describe things using explicit semantics. InChapter 4, we often said that the semantics of various OWL constructs implyother information. For example, a statement asserting that two classes arerelated usingrdfs:subClassOf implies that all members of the subclass arealso members of the superclass. Returning to the example in Chapter 4,141142 Part II ■ Foundations of Semantic Web ProgrammingDaisy is a member of the class ex:Canine because it is explicitly stated in thefollowing RDF excerpt:@prefix ex: <http://example.org/>.ex:Mammal rdf:type owl:Class.# Canine is a subclass of Mammalex:Canine rdf:type owl:Class;rdfs:subClassOf ex:Mammal.# Daisy is implicitly a member of the class Mammalex:Daisy rdf:type ex:Canine.The semantics of the rdfs:subClassOf relationship between ex:Canine andex:Mammal implies that Daisy is also a member of the class ex:Mammal.Inthisexample, the fact that Daisy is a canine is explicit. The fact that she is also amammal is implicit, or implied by the semantics of the ontology.OWL is merely an ontology language; it is not an application. As such,OWL alone doesn’t really do anything. OWL is a tool for specifying semanticsand defining knowledge models. Real-world applications must implement thefeatures of OWL to utilize the descriptive power of its semantics. To providethese capabilities, many Semantic Web applications use a framework of inte-grated components to provide the storage and retrieval of RDF information,as well as the interpretation of OWL semantics. We have already touched onthis subject in the introduction to this book, but now we will go into depthexploring these frameworks and their components.Semantic Web frameworks are used for many different purposes, includingdatabase translation and integration, domain knowledge modeling, validation,analysis, and even simply the storage and retrieval of information. Regardlessof their purpose, most frameworks provide the ability to create and manipulatea knowledgebase. A knowledgebase is a software component that representsa collection of information that is ontologically described, processed, andaccessed in a Semantic Web application. To provide this capability, a frame-work is composed of a set of tools, including an RDF store (often referred to as atriple store or graph store), an access API or query processor,andareasoning engine(or reasoner). Each of these components plays a critical role in providing thestorage and retrieval of RDF data, as well as the interpretation of the semanticsof OWL ontologies and instance data.While many Semantic Web frameworks are developed and released as com-plete packages, they usually allow developers to customize each componentto optimize the framework for a specific set of requirements. For example,the Jena Semantic Web Framework, which we will explore in greater depththroughout this chapter and the rest of this book, provides interfaces for you toincorporate your own RDF storage mechanism or your own reasoning engine.Chapter 5 ■ Modeling Knowledge in the Real World 143The Sesame RDF Framework provides a stack-based interface model thatallows you to insert your own layers, providing custom storage and inference.Developers have a large number of component implementations to choosefrom when customizing these frameworks because the components of theSemantic Web are built upon open standard interfaces, languages, and proto-cols. The following subsections explore the concepts, technologies, and toolsnecessary to apply Semantic Web technology to problems in the real world,beginning with Semantic Web frameworks.Semantic Web FrameworksMost Semantic Web frameworks are a collection of integrated tools that allowyou to create and work with a knowledgebase. The framework is the set oftools; the knowledgebase is the capability or concept of what they achieve.These frameworks are usually composed of three basic kinds of components,as depicted in Figure 5-1: storage, access, and inference. Each element isinterconnected because there is often a lot of interaction among these variouscomponents. Storage components are repositories of RDF statements that storeinformation. Access components are usually query processors or applicationprogramming interfaces (APIs) that provide the retrieval and modificationof information, and inference components are reasoning engines that applyinterpretation of OWL semantics to the information in the knowledgebase.RDF StoreBackward Chaining Forward ChainingReasonerQuery Endpoint APIsSWRLSTORAGEINFERENCE ACCESSRDF OWL SPARQLFigure 5-1 A Semantic Web framework (and knowledgebases in general) usually consistsof three fundamental components: storage, access, and inference.Fundamentally, a knowledgebase is a collection of facts (statements). Thecomponents of a Semantic Web framework serve to store, provide access to,and infer about these facts. Facts can be explicit or implicit. Explicit factsare those that have been directly asserted in the knowledgebase. Implicitfacts are entailments, facts whose existence is implied by the combination of144 Part II ■ Foundations of Semantic Web Programmingexplicit facts and the semantics of ontologies and rules in the knowledgebase.Entailments are derived by the reasoning component of the knowledgebase.Depending on implementation, entailments may be stored directly in theunderlying storage mechanism or they may be derived as needed when infor-mation is retrieved from the knowledgebase. Knowledgebase implementationsmay perform inference automatically or not, and they can perform inferenceinternally within the knowledgebase or through an external process.The modular design of most Semantic