180 likes | 297 Views
Concept Modeling on Semantic Wiki. Presented by Jie Bao, RPI baojie@cs.rpi.edu http://baojie.org Joint work with Li Ding and Zhenning Shuangguan Nov 22, 2008. SMW and me. For research TW group portal wiki-master
E N D
Concept Modeling on Semantic Wiki Presented by Jie Bao, RPI baojie@cs.rpi.edu http://baojie.org Joint work with Li Ding and Zhenning Shuangguan Nov 22, 2008
SMW and me • For research • TW group portal wiki-master • Testbed for ideas: semantic history, privacy-preserving reasoning, integrity constraint, close world reasoning • For fun and daily life • Blog • Time management (To-do list, issue tracker, calender) • Document everything (emails, papers, drafts,…)
Outline • 1. A formalization of SMW • The modeling language (SMW-ML) • The query language (SMW-QL) • 2. Template-based Rule Modeling • 3. OWL Meta Modeling (will be presented by Zhenning Shuangguan)
1. Formalization of SMW Why it matters? It • Gives a better understanding on SMW’s • potentials (what it can do) and • limitations (what it can not do) • Allows evaluation of implementation, e.g., soundness and completeness • Promotes community-wide adoption, e.g., on multiple DB platforms.
Modeling language (SMW-ML) • Syntax
Modeling language (SMW-ML) • Semantics: a variation of the RDF semantics • Resources can only be local wiki pages • N-ary property supported • No domain, range, blank node, collection and list • Classes and properties are disjoint • Allow punning between classes and individuals • Details in [Bao & Ding, 2008]
Modeling language (SMW-ML) • Entailment rules (sound and complete)
Query Language (SMW-QL) • Syntax in BNF
Query Language (SMW-QL) • Some observations • Subquery can allow “punning” of class & ind., e.g., • [[Has topic::<q>[[Category:Wiki]]</q>]] • [[Has topic::Category:Wiki]] • Selection: • top level is in Conjunctive Normal Form (CNF), e.g., (A and B) or (C and D) • “Concept” pages: • Are wiki “views” – may be tricky, e.g., concept containment, or cyclic definition.
Query Language (SMW-QL) • Semantics: by translating into logic programs (LP) • E.g., Husband(x) :- Man(x), marriesTo(x,y) • Why doing so? • Have a precise understanding of what a query means • Formal foundation for query answering, e.g., will the answer set always be unique? • Better position for future extensions, e.g., adding classical negation and negation-as-failure.
Query Language (SMW-QL) • Semantics: by translating into logic programs (LP) • Minimal Herbrand model semantics • Closed world assumption (why?) • Details in [Bao & Ding, 2008] • Example {{#ask: [[Category:Graduate Student]] [[Has affiliation::RPI]] |?Hobby }} Q(x,y) :- L(x), Hobby(x,y); L(x) :- A1(x), A2(x); A1(x) :- Graduate Student(x); A2(x) :- Has affliation(x,RPI);
2. Rule Modeling • Example: Template-based “Domain” inference {{#vardefine:value|{{#ask: [[:{{FULLPAGENAME}}]] |?{{{1}}}= |mainlabel=- |format=list |link=none }} }} {{#if:{{#var:value}}|[[Category:{{{2}}}]]}} Usage: {{Rule:Domain|hasAuthor|Document}} “Rule:Entailment”
2. Rule Modeling: Example http://tw.rpi.edu/wiki/Family_Rule
2. Rule Modeling • Simplicity – rule as page, easy to design, easy to use • Scalability • Each page can selectively turn on/off rule inference • Do not enforce completeness (which does not always matter on wiki) • Flexibility • Client-sided controllable • Highly extensible (rules can be added on the fly) • Expressivity: way beyond OWL. • Yet always terminates (due to incompleteness)
Ongoing work • Design Patterns on SMW: relational modeling, OO-modeling, advanced rule modeling • Help to bridge the gap between non-SW developer and SW technologies, • On-the-fly, client side expressivity extensions to basic semantic wikis, e.g., OWL-RL/SWRL/RIF support • Ontology Meta Modeling on SMW • OWL and OWL 2 • First Order Logic
Questions? Reference • Jie Bao and Li Ding (2008) Knowledge Representation and Query in Semantic Wikis: A Formal Study. Tetherless World Constellation Technical Report. http://www.cs.rpi.edu/~baojie/x/smw-bao-tr.pdf
Query Language (SMW-QL) Semantics: by translating into logic programs (LP)