220 likes | 234 Vues
Operationalizing the Geospatial Semantic Web. Spatial KR: Modeling in Practice. Also. Yaser Bishr, Ph.D. CTO, Co-Founder Image Matters LLC yasreb@imagem.cc. About this Talk. SHRDLU. The history of AI is full of examples of two weaknesses: Over-promising by insiders
E N D
Operationalizing the Geospatial Semantic Web Spatial KR: Modeling in Practice Also Yaser Bishr, Ph.D. CTO, Co-Founder Image Matters LLC yasreb@imagem.cc
About this Talk SHRDLU • The history of AI is full of examples of two weaknesses: • Over-promising by insiders • Over-optimism by outsiders • 35 years ago Ed Feigenbaum described Terry Winograd’s work as “a breakthrough in enthusiasm” • We surely don’t want this to happen to the Semantic Web • My goal is to outline what we need to operationalize the Geospatial Semantic Web • Talking points • Our approach at IM • Person: PICK UP A BIG RED BLOCK. • Computer: OK. (does it) • Person: FIND A BLOCK WHICH IS TALLER THAN THE ONE YOU ARE HOLDING AND PUT IT INTO THE BOX. • Computer: OK. (does it) • Person: WHAT DOES THE BOX CONTAIN? • Computer: THE BLUE PYRAMID AND THE BLUE BLOCK. • Person: WHAT IS THE PYRAMID SUPPORTED BY? • Computer: THE BOX.
Level Setting • An ontology specifies a rich description of the terminology, concepts, nomenclature • Properties explicitly defining concepts and relations among concepts (hierarchical and lattice) • Rules distinguishing concepts, refining definitions and relations (constraints, restrictions, regular expressions) • Relevant to a particular domain or area of interest.
Know the Big Picture Teaching, explaining Thoughts Ideas, Comments, Critics Enterprise Knowledge Means for sharing explicit knowledge Means for sharing implicit & tacit knowledge Working together to achieve a shared goal
Know What You Want To Tacit knowledge Explicit knowledge From Socialization Sympathized Knowledge Externalization KnowledgeEngineering Tacit knowledge Articulate tacit knowledge explicitly: metaphors, concepts, hypotheses, models Share experiences to create tacit knowledge. Combination Systemic Knowledge Internalization Operational Knowledge Explicit knowledge Learning by doing, to develop shared mental models and technical know-how. Manipulating explicit knowledge by sorting, adding, combining Nonaka & Takeuchi 1995: The Knowledge Creating Company
Characterize Your Application • Business Applications Hence Controlled Ontology • Transcendent source of structure, a high degree of formality, largely read-only. • External control relative to nearly all users and slowly changing • Analytic Applications Hence Dynamic Ontology • Wide ranges of authoritativeness, internally defined and highly changeable • Dynamic, flexible ontologies, using mostly large, read-only collections with a real-time focus. • Engineering Application Hence Standard Ontology • Wide ranges of authoritativeness, externally defined with volatile instances • primarily internally control with a focus on design-time application.
Manage Evolution But Do Not Control Ontology Personal • Ontology: Partial formalization of Conceptualization • Conceptualization is something personal you can’t tell me how I should see the world • Technology should allow personal/shared ontology Domain Personal Domain Domain Domain Upper-Level Ontology Domain Personal Personal Personal Domain Doman Personal
Determine Level of Control country cities capital Country Family UK Travel London Bridge with community knowledge Controlled No structure
Look Beyond OWL: Problem Solving Features Obs. Facilities Tracks Process Models Domain Models Spatial Models
Understand Knowledge Cube Knowledge with Pedigree and Provenance User Preferences and Action Core Business Process and State Ontology Tasks History Core Feature Ontologies Domain Processes and Tasks Core Geospatial, Topological, and Temporal Ontologies Domain Ontologies and Rules Geospatial Data and Metadata
Know That You Can Share More Than Tables SOA + SW Business Logic SOA KB XML Relational Data Stored Procedure DB triggers WS WS Triggers Stored procedure business logic Triggers Stored procedure business logic A380 wingspan 340 feet A380 can land on runway that has width of >= 150 feet If BA560 is A380 and Runway is R1 and is Approaching R1 then alert Air traffic control KB A380 wingspan 120 A380 can land on runway that has width of >= 150 feet If BA560 is A380 and Runway is R1 then alert Air traffic control A380 wingspan 120’
Because you are sharing knowledge, you Still Have More Hurdles
Know Your World SOA + SW Is BA560 only plane landed on R1? OpenWorld: I don’t know ClosedWorld: Yes Business Logic Relational Data Stored Procedure DB triggers WS A380 wingspan 120 A380 can land on runway that has width of >= 150 feet If BA560 is A380 and Runway is R1 and is Approaching R1 then alert Air traffic control BA560 Landed on R1
Integrate, Don’t Migrate MySQL Oracle Semantic Data Store WFS SDE • Build ontologies with an eye on performance bottlenecks of reasoners • Build ontologies knowing what you will use them for • Commercial new RDF databases not a good idea • Let DB vendors do what they do best, Oracle RDF data store: (concurrency, scalability, performance, security, etc.) • Legacy databases also include stored procedures and triggers • You are building a KB layer not a database view • RDF stores will host the KB • They will reference legacy databases Semantic Web Knowledge Integration Coverage ShpFiles Data Sources
Legacy Vs Semantic Data Stores Semantic Data Store Notify security when car approaches CourtHouse Select Highways within bounding box Car Approach CourtHouse HW=Road(Speedlimit >=105 KM/h and Lanes>=3) Car IsA MobileObject IsA Feature CourtHouse isA Building IsA Feature Road and Highway IsA Feature Approach hasType (Distance@T2 < Distance@T1) MobileObject Approach Feature Semantic Data Integration Shapefile Oracle
For Geospatial We Need • The Easy ones • What Jerry talked about + • Frame of reference • A Behind B, X right behind A, then X is in front of B • Orientation • What if A and B have their backs to each other, then X is also behind B • Identity • What are the characteristics that make distinguish a lack from a pond • Optional Vs mandatory semantics: is it because I say so, or because they determine identity • Is it RCC8 or 9-interseciton or both? • Moving Objects • Approaches, going farther, ascend, descend, defining constraints of moving objects wrt to each other and wrt fixed objects (road network) • Spatial data types • Spatial Spatial SQL a la Spatial SQL • The Hard ones • Geospatial Boundaries (man made Vs natural) • What is a Features • Domain ontologies
OWL should be less Verbose • Flight = Airplane Π >1 FlightNumber <owl:Class rdf:ID=“Flight"> <owl:intersectionOf rdf:parsetype="Collection"> <owl:Class rdfs:about=“Airplane" /> <owl:Restriction> <owl:onProperty rdf:resource=“FlightNumber" /> <owl:minCardinality rdfs:datatype="&xsd;Integer"> 1 </owl:minCardinality> </owl:Restriction> <owl:intersectionOf> </owl:Class>
Ontologies Must be Managed as Part of Enterprise Architecture • OMG has defined Ontology Definition Metamodel • A standard meta-model for ontology modeling • Part of MDA • A UML2 Profile for depicting Ontologies • With at least mappings • Between ODM and the profile • Between ODM and the W3C OWL DL
Bridging KR and MDA From ODM slides at OMG • We need composable ontologies. I don’t need whole sale OWL “imports” • We need version control • We need ontology to be part of configuration management • We need to minimize the learning curve: learn from Google Ajax approach
Geospatial Knowledge Engineering LoB at IM SEMANTICSTECHNOLOGY SYSTEMSENGINEERING METHODOLOGY GEOSPATIAL ASPECT SCICOP and COI Presence OGC US Army: Semantic Mapping Tools AFRL: Semantic Roadmap NGA: Problem Responsive Data Discovery and Retrieval NGA: Automated Information Triage for Deep Geospatial-Intelligence Analysis DOMAINS Enabling solution: knowledgeSmarts
knowledgeSmarts™ MySQL Semantic Data Store Oracle WFS SDE Triage Works Application Interface Triage Knowledge Model Lenses KS Engines DL Process Spatial Temporal CBR Domain Models ECA KMS Query Engine Query Optimizer Query Builder Distributed Query Manager ksMapper Coverage ShpFiles Data Sources
Thank You yaserb@imagem.cc