Reasoning Web 2007 summer school Ondej vb UEP
Reasoning Web 2007 summer school Ondřej Šváb (UEP)
Summer school • • 3 rd summer school of Reasoning Web Malta, Lisbon, Dresden Around 40 participants (canditates of Ph. D), Many contributors to tutorials • LNCS Proceeding – ¨Fundamentals of Reasoning and Reasoning Languages¨ • query paradigms, declarative semantics, complexity and expressive • power, procedural semantics, implementations, optimisation, and • many more. . – ¨Rules and Policies¨ – ¨Applications of Semantic Web Reasoning¨ • Supported by the European Network of Excellence (REWERSE)
¨Fundamentals of Reasoning and Reasoning Languages¨ • Francois Bry et al. • Foundations of logic – First-Order Predicate Logic (syntax, Tarski Model Theory, Herbrand Model Theory) – Tarski inadequacies • • Unique name assumption Terms as compound data structures Closed world assumption Disregard infinite models
¨Fundamentals of Reasoning and Reasoning Languages¨ • Francois Bry et al. • Foundations of logic – First-Order Predicate Logic (syntax, Tarski Model Theory, Herbrand Model Theory) – Tarski inadequacies • • Unique name assumption (Herbrand models) Terms as compound data structures (Herbrand m. ) Closed world assumption (minimal Herbrand m. ) Disregard infinite models (finite model theory)
Declarative semantics • Minimal Model Semantics of Definite Rules (Herbrand models) • Fixpoint Semantics of Positive Definite Rules • Declarative Semantics of Rules with Negation (CWA) • …
Operational semantics • Naive evaluation of datalog programs – Extensional predicates are mapped to intensional predicates • SLD-resolution: basic backward chaining – SLD=Selected Literal Definite Clause – SLD-tree – disadantage: generally infinite • OLDT-resolution – tabling – store intermediate results for later usage – Better performance, but still non-deterministic alg.
Operational Semantics • The magic templates transformation alg. – Both, goal-directed search as well as forward chaining – 2 steps: • Transformation into adorned version (sip) • Rewriting adorned version into set of rules that can be efficiently evaluated
Operational Semantics • Rete algorithm – By Charles Forgy in 1974 – alg. For inference engines of rule based ES – eg. in Clips, JRules, OPS 5 – Two types of memories: • Production memory (rules) • Working memory (ground facts, implied facts) • Processing of rules in Rete network (four types of nodes) – Forward Chaining, sharing information
Rete algorithm
Complexity and Expressive Power of Logic Programming Formalisms • Nice introduction into complexity (turing machine, complexity classes) and much more
Reasoning in Description Logics by Ulrike Sattler • The logic basis for OWL flavours • ‘common’ reasoning in service based systems for ontology developing – – The computation of the taxonomy Checking the satisfiability, consistency Satisfiability and subsumption wrt. TBox Consistency and instance wrt. ABox • Others (non-standard inferences services): – – Debugging support to explain unintended specialisation links Extraction of sub-ontology describing the concept Find (the most specific) generalisation Find approximation of a concept
DLs • Syntax, semantics, reasoning • Many exercises: interpretations, tableaux algorithm (satisfiability of concept) – Pellet, Racer. Pro, Fa. CT++ • Other reasoning approaches: – Disjunctive datalog engine (KAON 2) – Technique from deductive databases – Rule-based (rule engine) (OWLIM), costly materialization strategy • DLs and rules (combination) – Reasons for rules, p. 177 • Adding non-monotonicity (rules have CWA, negation as failure) • Generally increasing DLs’ expressive power
¨Rules and Policies¨ • Reactive rules on the Web – Rule based reactive systems on the web – Two approaches • Event-Condition-Action rules – ON event IF condition DO action • Production rules – WHEN condition DO action – Examples of application scenarios – Examples of languages – Differences…
¨Rules and Policies¨ • Rule-based policies – Security policies, trust mng policies, business rules, … – Requirements for a policy framework – Examples of languages KAOS and Protune
Rule Interchange on the Web • Standardisation effort of RIF WG within W 3 C • Not trivial task – different current rule languages (diverse syntax and semantics) – Deduction, normative, reactive rules • So far RIF WG – Uses cases and requirements (implementability, semantic precision, extensible format, translators, standard components)
RIF • RIF core – Horn rules • RIF dialects - extensions
¨Applications of Semantic Web Reasoning¨ • The contradiction of Web and Reasoning • Reasoning in Semantic Wikis – Reasoning for enhancing browsing, querying, editing, validating – Wikis for semantic data (collaborative ontology editors) – Semantic data for wikis (simplifying maintenance, retrieval, …)
Semantic Media Wiki • Extension of Media. Wiki • Many ABox statements, shallow TBox • Annotations made by people via a special markup (categories, relations, attributes)
Ike. Wiki • In contrast to SMW, preloaded ontologies • „restrive approach“ wrt. Annotation • Reasoning – subclass and type inference
Examples of reasoning now • Displaying and browsing wiki contents – Graphical visualisation, … • Editing support – suitable properties (domain), values wrt. Consistency • Validating against background knowledge
SMW – complex inferencing? • Challenges: – maintaining ontologies in wikis (axioms vs. articles), higher expressivity – Performance – semantic query answering for large audience (semantic Wikipedia) – Usability – improved visual interfaces – Reasoning about context (suitable diagram) – Planning support (resource planning, appropriate date and people) – robustness
¨Applications of Semantic Web Reasoning¨ • Reasoning in Semantic Wikis (AIFB, Salzburg r. ) – Not much reasoning, ideas • Semantic Web Service Discovery (SAP r. ) – Subsumption, request, offer • Intergator – information retrieval mng system – GATE, Jena, Lucene • Knowledge based search technologies – Biomedical literature, www. gopubmed. org – Searching by browsing over structural results
- Slides: 23