2426 Semantics for SDI

Semantics for Spatial Data Infrastructures : 

Semantics for Spatial Data Infrastructures Werner Kuhn

Slide2: 

The Economist, May 3, 2003

Slide3: 

The Economist, May 17, 2003

Slide4: 

Topological Operators for areas, lines, points

Slide5: 

Overlap in a GIS dialog window

Slide8: 

Composite Service

Slide9: 

Semantic Heterogeneities

Generalized Problem: 

Generalized Problem Buffers, topology, navigation, and gas plume examples show problems with the semantics of geographic information describing entities (e.g., road, ferry) processes (e.g., driving, wind) relations (e.g., distance, overlap) Context is essential for semantics Today‘s SDI separate data from operations (GML) context contained in operations is lost systems or users misinterpret data Ontologies are supposed to provide that context Today, they don‘t They could do better How?

What do we mean by Semantics ?: 

What do we mean by Semantics ? Jaguar Referent Symbol stands for refers to activates Concept

Semantics of what ? : 

Semantics of what ? We are not trying to capture the meaning of natural language expressions! We formalize the semantics of technical symbols used in GIS, databases, web services of information communities But: information is from and for humans it derives from data through human interpretation it emerges at the user interfaces of GI technologies Thus, we need to capture meaning (cognitive semantics)

Medium-term research program(3-5 years): 

Medium-term research program (3-5 years) solve 3 semantic interoperability problems: e.g., road data for directions e.g., wind direction for gas plume e.g., weather services data discovery service discovery service composition

Semantic Interoperability : 

Semantic Interoperability ...is the only real interoperability interoperating components share an understanding of their interfaces today: „syntactic interoperability“ interoperating components share an interface, defined by a type signature GM_Object :: distance (geometry : GM_Object) : Distance

Today: annotation: 

Today: annotation ISO 19107, Spatial Schema „The operation "distance" shall return the distance between this GM_Object and another GM_Object. This distance is defined to be the greatest lower bound of the set of distances between all pairs of points that include one each from each of the two GM_Objects. A "distance" value shall be a non-negative number associated to a distance unit such as meter or standard foot. If necessary, the second geometric object shall be transformed into the same coordinate reference system as the first before the distance is calculated. If the geometric objects overlap, or touch, then their distance apart shall be zero. Some current implementations use a "negative" distance for such cases, but the approach is neither consistent between implementations, nor theoretically viable. "Distance" is one of the units of measure data types defined in ISO TS 19103. NOTE The role of the reference system in distance calculations is important. Generally, there are at least three types of distances that may be defined between points (and therefore between geometric objects): map distance, geodesic distance, and terrain distance.“

Goal: axiomatization: 

Goal: axiomatization Service interfaces, requests, and responses contain symbols with undefined semantics GM_Object :: distance (geometry : GM_Object) : Distance Alexandria :: distanceStades (Syene) = 5040 type symbols (standing for classes of objects and literals) values (standing for individual objects and literals) operators (standing for methods)

Ontology in Philosophy: 

a philosophical discipline—a branch of philosophy that deals with the nature and the organisation of reality Science of Being (Aristotle, Metaphysics, IV, 1), asks the questions: What exists? What characterizes being? Eventually, what is being? Ontology in Philosophy

Ontology in Computer Science: 

Ontology in Computer Science An ontology is an engineering artifact. It is constituted by a specific vocabulary used to describe a domain assertions on the intended meaning of the vocabulary. Thus, an ontology describes a formal specification of a certain domain: shared understanding of a domain of interest formal and machine manipulable model of a domain “An explicit specification of a conceptualization” [Gruber93]

Ontologies: 

Ontologies

clc: CORINE Land Cover categories: 

clc: CORINE Land Cover categories

Are ontologies enough?: 

Are ontologies enough? They contain relationships between terms superclass and subclass, is_a, part_of, has_property, synonym Their axioms are absent or in obscure and untested logical axioms They fail to capture context activities determine semantics (e.g., driving across a river) They are not grounded what does „artifact“ or „entity“ or „process“ mean?

The underlying problem: 

The underlying problem geoinformation = < x, z > ???

Vision: 

Vision Users of geographic information should be able to refer thematic data to semantic reference systems, just as they refer geometric data to spatial reference systems. Software should support the referencing and grounding process projections to simpler semantic spaces semantic translation among different reference systems.

Semantic Reference Systems: 

Semantic Reference Systems Spatial Reference System [ISO 19112] „system for identifying positions in the real world“ e.g., geodetic reference systems Temporal Reference System [ISO 19108] „basis against which time is measured“ e.g., calendars Semantic Reference System [t.b.d.] basis on which thematic data are interpreted: ‚forest‘, ‚wetland‘‚ ‚road width‘ etc. e.g., ontologies

Long-term research program(5-7 years): 

Long-term research program (5-7 years) create methods and tools to design and use semantic reference systems for grounding (e.g. “move”) projecting (e.g. roads and ferries to edges) translating (e.g. cadastre to navigation) based on real-world case studies transportation, emergency management, planning

Ferry example: Grounding: 

Ferry example: Grounding

Grounding in Image Schemas : 

Grounding in Image Schemas Sensory-motor patterns of cognition (Johnson) container, surface, path, link, center-periphery, force... developed through bodily experience have internal structure enable the perception of meaningful information from the environment (Gärdenfors)

Ferry example: Semantic Projection: 

Ferry example: Semantic Projection

Gas plume example: Translation: 

Gas plume example: Translation

Semantic Reference System: 

Semantic Reference System

Protégé Implementation: 

Protégé Implementation

Problem: Semantic Heterogeneity 1: 

Which terms can I use for an efficient keyword-based search of water level information? Problem: Semantic Heterogeneity 1 Synonyms: one concept – different terms water level, tide scale measuring gauge, control point Homonyms: one term – different concepts „water level“ – river vs. groundwater  The returned results are incomplete!  Not all returned results are relevant! Discovery

Problem: Semantic Heterogeneity 2: 

Problem: Semantic Heterogeneity 2 How do I interpret the terms of a schema and formulate a query? Retrieval  Meaning of terms used in the database schema is often ambiguous

Approach: Ontological Descriptions: 

Approach: Ontological Descriptions What is the water level at point X at time Y in the Elbe River? Ontological description of the application concept “höhe” Application Ontology Concept

Architecture: Ontology-Based Discovery: 

Architecture: Ontology-Based Discovery

Architecture: Ontology-Based Retrieval: 

Architecture: Ontology-Based Retrieval Request for feature type Request for concept definitions GetFeature request

Why Spatial is Special (for Semantics): 

Why Spatial is Special (for Semantics) Space and time are primarily understood through processes Semantics of symbols is not always a matter of convention physical grounding needs measurement ontologies But human perception and social reality play an important role simplest case: geographic names and location descriptions Need spatial reasoning at the type level Granularity, vagueness, uncertainty and are central aspects of spatial information ...

Conclusions: 

Conclusions Spatial reference systems motivate semantic reference systems from static ontologies to computational reference systems providing referencing, projection, and translation Semantics of spatial information poses both, general and special challenges. Our success measures are in solving interoperability problems

Some Big Challenges: 

Some Big Challenges Grounding of ontologies („semantic datum“) what is meant by wind direction? what is transportation? Reasoning about processes what does distance mean? what does touch or overlap mean? what is driving? how is a gas plume determined? Improved similarity models ad hoc categories Karten bilden bei logistischen Entscheidungen der Helfer eine wichtige Grundlage: "Muss ich einen Helikopter einsetzen oder kann ich fahren?" Auch bei der Einrichtung von Hilfsstationen seien Karten und Bilder wichtig: "Wo habe ich eine ebene Fläche, die einigermaßen weit weg vom Wasser ist, die groß genug ist, um ein Zelt-Lazarett aufzubauen?“ [dpa Gespräch mit Robert Meisner, DLR, 7. Januar 2005] Formal Definition of Interoperability Automated generation of semantic annotations

For more information...: 

For more information... MUSIL web site (Muenster Semantic Interoperability Lab): http://musil.uni-muenster.de email to kuhn@uni-muenster.de