PowerPoint PresentationSlide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Slide 27Slide 28Slide 29Slide 30Slide 31Slide 32Slide 33Slide 34Slide 35Slide 36Slide 37Slide 38Slide 39Slide 40Slide 41Slide 42Slide 43Slide 44Slide 45Slide 46Slide 47Slide 48Slide 49Slide 50Slide 51Slide 52Slide 53Slide 54Slide 55Slide 56Slide 57Slide 58Slide 59Slide 60Slide 61Slide 62Slide 63Slide 64Slide 65Slide 66Slide 67Slide 68Slide 69Slide 70Slide 71Slide 72Slide 731Semantic Web FrameworksSlides Created By Jeff Partyka, taken from the book, “Semantic Web Programming”2Lecture Outline1. About Semantic Web Frameworks2. A Glimpse At RDF Stores3. A Glimpse At Reasoners4. Common Frameworks and Components5. OWL Profiles6. Comparing Inference in RDFS and OWL7. Ontologies in the Real WorldSo for most of the course so far….•We’ve learned about RDF, RDF Schema and OWL•These are semantic data modelling languages, when it comes down to it•By themselves, these don’t actually do anythingSo how do we use these languages?•We need a Semantic Web Framework to utilize the semantics encoded in these languages•We combine the semantics with different components of the framework, like RDF stores, reasoners, a SPARQL query processor, and a programming language API (like Jena)A Model of a Semantic Web Framework•The basic components of a framework interact with each other throughout the lifecycle of a Semantic Web application:A Model of a Semantic Web Framework(cont)•Semantic Web Frameworks are composed of 3 kinds of components: (1: storage (2: access (3: inference•The framework is set of tools, the knowledgebase is the capability or concept of what they achieve•A knowledgebase is a collection of facts•Components of a semantic web framework serve to store, provide access to and infer about these factsKnowledgebases and Facts•Facts can be explicit or implicit•Explicit facts are those that have been directly asserted in the knowledgebase•Implicit facts are entailments, facts whose existence is implied by the combination of explicit facts and the semantics of ontologies and rules in the knowledgebaseEntailments•Entailments are derived by the reasoning component of the knowledgebase•Depending upon the implementation, entailments may be stored directly in the underlying storage mechanism, or they may be derived as needed when information is retrieved from the knowledgebaseCustomizing Semantic Web Frameworks•Semantic Web frameworks have a modular design, allowing for a user to plug in custom components for a specific set of requirements•For example, if the application requires a very fast system that can operate over large volumes of data with minimal support for OWL semantics, you can plug in (1: a scalable, persistent RDF store (2: a minimal inference componentCustomizing Semantic Web Frameworks (cont)•If a requirement called for complete OWL inference capabilities without much concern for scalability and large volumes of data, a very lightweight in-memory RDF store could be used in combination with a powerful reasoning engineStoring and Retrieving RDF•Semantic Web frameworks require storage and access mechanisms•However, the type of storage depends on the use case•An RDF store can range from a small in-memory model with a retrieval API to a server-based RDF store capable of storing billions of statements with a query processor that can handle hundreds of concurrent connectionsRDF Store Implementations•These persist triples for a given RDF graph•The Subject-Predicate-Object model of RDF does not mesh well with traditional relational DB storage models•This is because relational models rely on tables with a static number of attributes for all tuples, and where each attribute possesses a data type which describes the possible range of the dataExample: Relational DB triple storesProblems Storing RDF in RDB tables•In the previous slide, a single table consisting of the attributes “subject”, “predicate”, “object” is used to hold all RDF triples over one or more graphs•This is done because (1: the triples of RDF graphs tend to span multiple domains, making it more difficult to create tables representing common entities (2: Semantic Web data is inherently distributedProblems Storing RDF in RDB tables (cont)•However, this approach does not scale well, because most real-world queries are complex and involve multiple joins•Having all RDF data stored in one table would guarantee that EVERY query would require several joinsExample query for RDF in RDB query languageThis query retrieves all authors of books whose title contains the word “Transaction”Graph databases•Graph databases have been used to more accurately model the structure of RDF data and mitigate some of the performance problems of relational model based stores•They make heavy use of locality•For example, for 2 statements that share the subject, they will be stored close to each other on disk.Graph databases (cont)•Graph databases link statements sharing common resources in a linked list, in the same way that we represent RDF graphs•Some examples of graph databases: HypergraphDB, Neo4j, AllegroGraph, OpenLink Virtuoso, OpenLinkDBRealizing the Semantics of OWL•A semantic web framework applies a reasoner to the triples in an RDF store to entail new triples•Entailment can occur via a rules engine, database triggers, decision trees or even using programmatic business logic•Rule-based reasoning engines combine the statements within an RDF store with a set of logical rules in order to derive assertions or perform actionsRule-based Reasoning•A rule comprises of two parts: an antecedent and a consequent. A rule models an if-then statement•Rules can be used to express OWL semantics, as well expressing statements that cannot be modeled in OWL•In a rule, any time the statements in the antecedent are true, the statements in the consequent can be inferredRule Example #1Rule Example #2SWRL Rules•The previous rules are an example from a rule language. The de facto Semantic Web Rule Language is called SWRL•SWRL rules allow for more expressive inferences that are impossible in OWL•Limitations of OWL: No built-in functions, no support for (1: string transformations (2: pattern matching (3: datatype and unit conversions(4: Certain types of math