Computer Languages and Computing Science Department University of Malaga Spain A Survey on Disk Oriented Querying and Reasoning on the Semantic Web María del Mar Roldán , José F. Aldana Spanish junior researchers fight for employment rights http://khaos.uma.es We need (efficient and scalable) TOOLS for storing and querying OWL ontologies Databases DL Systems Integration not trivial Study how different logical and physical (persistent) storage models influence on the querying and reasoning mechanisms performance 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 2 Objectives • Describing seven systems (SESAME, JENA, KAON, KOWARI, DLDB, DLP-IM, IS) which uses database technology to both represent knowledge persistently and query this knowledge in the Semantic Web context. • According to four features • Ontology definition language • Storage model • Query language • Reasoning mechanisms • Obtaining conclusions and needs 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 3 Evaluation Framework (I) 1. Ontology Definition Language 1.1 Language used (expressiveness for representing knowledge) 2.1 Storage Technology 2. Storage Model (Tbox and Abox Storage) 2.2 Specific physical representation 2.3 Access paths (idexes) 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 4 Evaluation Framework (II) 3.1 Type of queries allowed (language expressiveness) 3. Query language 3.2 Scalability and efficiency in massive storage 3.3 Distribution 4.1 Tbox reasoning mechanisms 4. Reasoning Mechanisms 4.2 Abox reasoning mechanisms 4.3 reasoning mechanisms implementation 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 5 Ontology Definition Language • From RDF to OWL SESAME JENA KAON KOWARI DLDB DLP-IM IS RDF RDF RDF + extensions = Part of OWL expressiveness. RDF OWL OWL subset OWL 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 6 Storage Model • Databases without specific physical representation and access paths. SESAME JENA KAON KOWARI DLDB DLP-IM IS Object Relational DB + Relational DB Relational DB Relational DB Java database + AVL indexes Relational DB Deductive Database Relational DB 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 7 Query language • Usually translated to SQL or Datalog • No distribution SESAME JENA KAON KOWARI DLDB DLP-IM IS RDQL API SeRQL API RDQL API KAONQL Datalog iTQL OWL SQL Datalog OWL SQL 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 8 Reasoning Mechanisms (Tbox) • Concept subsumption • No implementation homogeneity SESAME JENA KAON KOWARI DLDB DLP-IM IS Concept Subsumption Concept Subsumption Those that can be translated to Datalog Concept Subsumption Fact mechanisms Those that can be translated to Datalog NO 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 9 Reasoning Mechanisms (Abox) • Instances Retrieval • No implementation homogeneity SESAME JENA KAON KOWARI DLDB DLP-IM IS IR + domain/range of properties IR Generic rule reasoner + incomplete owl lite reasoner (JENA2) Transitive and symmetric properties IR IR Those that can be implemented by Datalog rules IR 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 10 Current tools (Using database technology) Databases as instances warehouses Our Proposal Select the best logical model Simple logical models No physical representation of the knowledge Select the best physical storage structures and indexes Few reasoning mechanisms (subsumption) and simple query languages Implement complex reasoning mechanisms and query language 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 11 Current tools (Using database technology) Databases as instances warehouses Select the best logical model Simple logical models No physical representation of the knowledge Select the best physical storage structures and indexes Few reasoning mechanisms (subsumption) and simple query languages Implement complex reasoning mechanisms and query language In general (for all OWL ontologies) Methodology For some specific applications (biology) 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 12 How to proceed (Future tends) • Defining complex reasoning mechanisms • Defining our query language: complex concepts + conjunctive queries • Implementing the reasoning mechanisms and the query language using several persistent storage models (relational databases, logical databases, XML databases…) 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 13 How to proceed (Future tends) • Designing different physical storage for the previous models • Getting performance measures and doing analysis • Obtaining conclusions and using them to define a set of recommendations/methodology 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 14 Problems…. • Technical Problems: •Complexity of the language (DL formalism) •No good OWL parser • BIG QUESTIONS TO DEAL WITH: •Face OWA-CWA impedance mismatch in the query language •Analyze (New) Trade-offs between expressiveness and efficiency (when reasoning in secondary memory) 01/07/2016 Semantic Web and Databases 2006 María del Mar Roldán-Garcia 15 Computer Languages and Computing Science Department University of Malaga Spain A Survey on Disk Oriented Querying and Reasoning on the Semantic Web María del Mar Roldán , José F. Aldana Spanish junior researchers fight for employment rights http://khaos.uma.es