Nov 10, 2016 - XML physical schema. (terms fully defined ... First-Order Logic. Ontology. INSPIRE. HY_Features hydrontol
DOMAIN REFERENCE ONTOLOGIES VS. DOMAIN ONTOLOGIES: WHAT'S THE DIFFERENCE? Lessons from the Water Domain Torsten Hahmann, UMaine, NCGIA Shirly Stephen, UMaine Boyan Brodaric, GSC
OUTLINE 1. Motivation • The Water Domain • State-Of-the-Art in Hydro Ontologies • The Role of a Domain Reference Ontology 2. Use Case • Objective & Approach • Hydro Foundational Ontology (HyFO) • Groundwater Markup Language (GWML2) • Results 3. Takeaway
2
THE WATER DOMAIN
Water Features (surface, subsurface, atmospheric) & their location and spatial configuration (rock bodies, surfaces, depressions) & Flow between them
3
Scope & Level of Detail
Generality WaterML2.0 RiverML
Application-specific standard (for integrating water observations)
•
Application-specific standard
•
Hydro domain ontology
• •
HY_Features
hydrOntology Surface Water Ontology (NHD) Surface Water Schema
GWML
XML
Models geometry/morphology of rivers XML, builds on WaterML
OWL
• Hydro domain ontology
• • •
surface and subsurface domain groundwater model based on GeoSciML does not describe the concept of void
UML (terms fully defined only in a glossary)
Hydro domain ontology
• •
emphasis on surface water features does not contain and hydrologic relations
•
150 classes, 34 object properties, 66 data properties., 256 axioms concepts varying from global-continental to regional-local
Application ontology (for interoperability among data sources of IGN-E) Hydro domain ontology
• •
Hydro domain ontology
Application-specific (built using GeoSciML and O&M schemas)
Foundational Grounding
For time-indexed hydrologic observations
surface and sub-surface domains approximately 50 concepts with mostly taxonomic relations not rich in relations aspect
SWEET INSPIRE
Format
•
• •
Based on The National Hydrology Dataset in the National Map specifies taxonomic type of hydro features Distinguishes between container (solid object) and water body in the surface domain For the exchange of groundwater related information Contains hydrogeo units, voids and wells
UML
XML
OWL axioms RDF triples with a SPARQL endpoint
BFO
OWL DL
UML conceptual schema, XML physical schema (terms fully defined only in a glossary)
4
Existing ontologies for the water domain – and for most scientific domains – rely on conceptual models (UML diagrams) and lightweight ontologies (RDF or subclass hierarchies specified in OWL)
OWL Full or First-Order Logic Ontology
Semantic nuances in usage RDF or OWL DL Ontology of terms is not formally Semantic Water SWEET Quality Portal Extension Ontology expressable in these HIS Ontology hydrontology languages Conceptual
HyFO
Semantic Interoperability
NADM
Geoscience Terminology WG
Model
GWML NHD Ontology Petroleum Basin Ontology
HY_Features
Exploration Ontology INSPIRE
Schematic Interoperability
eMinerals
Controlled Vocabulary
RiverML HydroML
WaterML
GeoSciML
Terminological Interoperability
GML
HydroXC AGU Term Index
Syntactic Interoperability
5
Spring
Catchment Surface Water Body Tributary
Channel Watershed
Subsurface water Runoff
Aquiclude/Aquitards
Lake Well
River
Hydrogeo Unit Stream
Basin
Reservoir
Cloud
Groundwater
Hydro Unit 6
SEMANTIC NUANCES: WHAT IS AN AQUIFER? GWML2 A hydrogeological unit that potentially stores groundwater.
INSPIRE An aquifer is a rock body, but does not capture the notion of voids in it.
NWIS (USGS) Geological formation or structure that supplies water to wells/springs. These subtle differences are currently not formalized typically only available in narrative form prevents use of automated integration (e.g., ontology alignment) techniques
How do we reconcile these semantic differences?
7
HOW DO DOMAIN REFERENCE ONTOLOGIES HELP? Vision: Leveraging a domain reference ontology (here: the hydro reference ontology HyFO) to analyze and revise geoscience knowledge representations (here: hydro data models) to increase semantic interoperability among them.
Exhibits many characteristics of foundational ontologies, but more specific: they are foundational for their domain
8
WHAT DOES A DOMAIN REFERENCE ONTOLOGY LOOK LIKE? Exhibits many characteristics of foundational ontologies: foundational for their domain 1. 2.
3.
Foundationally grounded Broad coverage on the highest level in the domain: focuses on the key concepts and relations in the domain; but does not aim to capture the domain comprehensively Specified in a highly expressive and fully machine-interpretable ontology language Provides “neutral” language to express semantic differences; Purpose is not to directly define the scientific terms (e.g. aquifer), but ontological helper concepts and relations
Domain Coverage Foundational Grounding
Domain Reference Ontology
Concise Formalization
9
DEVELOPING DOMAIN REFERENCE ONTOLOGIES Significant time needs to be invested into their careful development • Identify key concepts and relations in the domain • Need to be developed by ontologists, though with input from domain experts • Shortcuts just create another domain ontology, but not a reference for the domain • Payoff: Once finished, can be used to analyze and improve multiple domain and application ontologies and increase their semantic interoperability
Foundational Ontology Generic Ontology e.g., Location,
Fluids
e.g., HyFO
e.g., DOLCE, GFO
Domain Reference Ontology
Domain Ontology
e.g., GWML, INSPIRE
Application Ontology
e.g., NHD
10
OUR USE CASE: OBJECTIVE & APPROACH We tested whether a domain reference ontology can be successfully used to help clarify the semantics and improve their formalization of existing domain ontologies Role of Domain Reference Ontology: provide ontological guidance and formal language to analyze other domain ontologies
Domain Reference Ontology: Hydro Foundational Ontology (HyFO) Analyzed Domain Ontology: Groundwater Markup Language, v2 (GWML2), specified as UML diagrams with English explanations Step 1: Formalize all aspects of GWML2 semantics in first-order logic using HyFO terminology
Step 2: Identify and resolve semantic ambiguities and inconsistencies in GWML2
Step 3: Verify that the resulting GWML2-FOL is a logically consistent extension of HyFO
11
HYDRO FOUNDATIONAL ONTOLOGY(HYFO) Under development (Hahmann, Brodaric, Gruninger) since 2011 Progress preseted in a number of papers (voids, containment relations, constitution/dependency, water features) Container Object
Void
The solid object where water can be located; e.g., the channel of a river, the rock body in an aquifer.
Container Solid Body (CSB)
hosts Hydro Void (HV) hosts contains
constituted-by
contains
Matter Material that constitutes a container or water object; e.g., solid rock matter, water matter.
Matter (M)
Space(s) in the container that can be filled with water; e.g., pores in an aquifer, depression in a channel.
constituted-by
Water Body (WB)
Water Object
The liquid object located in the container/void; e.g. the spatio-temporal object that encompasses all the water in an aquifer.
12
HYFO RELATIONS Spatial relations Spatial part Spatial overlap Different kinds of spatial connection
Physical relations: Material-spatial dependence Containment Constitution Hosting a void
Different kinds of physical parts
13
GROUNDWATER MARKUP LANGUAGE (GWML2) OpenGIS Consortium (OGC) standard for the exchange of groundwater information. Version 2 (GWML2) currently under development. UML schemas with narrative semantic descriptions.
Extends GeoSciML (a markup language for geosciences) to describe hydrogeological concepts. Extends OGC/ISO Observations and Measurements schema to describe concepts and properties relevant to flow of groundwater.
Earth Material
Vulnerability
FluidBodyProperty
Management Area
Hydrogeo Void
Porosity
UnitFluidProperty
Yield UnitVoidProperty
GWML2 CM - GW Properties Geologic Unit Fluid Body
FluidBodyChange
Hydrogeo Unit
UnitProperties
Constituent Relation
Mixture
Aquifer Unit
Basin
Constituent Fluid Body Surface Aquifer Chemical Constituent
Divide
Biologic Constituent
Material Constituent
GWML2 CM - Fluid Body
Confining Bed GWML2 CM Hydrogeological Unit
Aquifer System
UnitProperties
Flow
Interflow
Monitoring Site
Well GWML2 CM -Wells
Flow System
Intraflow
Spring Discharge
Recharge
GWML2 CM - Flow
14
RESULTS 1: FORMALIZATION OF GWML2 IN FIRST-ORDER LOGIC (GWML2-FOL) WITH CLARIFIED AND REFINED SEMANTICS USING HYFO’S TERMINOLOGY A HydroVoid that is hosted by the Solid Body that is a submaterial of a Hydrogeo Unit or a Well. HGV(x) ≡ HV(x) ∧ ∃y,z [SB(y) ∧ hosts-v(y,x) ∧ submat(y,z) ∧ (HGU(z) ∨ W(z)]
AWB(x) ≡ WB(x) ∧ ∃y,v [A(y) ∧ ∀z [intra-const(z,y) ∧ WM(z) → submat(z,x)] ∧ HGV(v) ∧ P(r(x),r(v)]
Hydrogeo Void
contains
hosts-v Specialization of Matter
constituted-of
Aquifer Water Body
Aquifer
contains
EM(x) → M(x)
Earth Matter
Water Body that consists of all water in an Aquifer and resides in a Hydrogeo Unit.
Solid Body
A NAPO that is constituted of non-fluid Earth Material. SB(x) ≡ NAPO(x) ∧ ∃y[M(y) ∧ intra-const(y,x) ∧ ∀z (intra-const(z,x) → ¬FM(z))]
subclass-of
Hydrogeo Unit A specialization of Hydro Rock Body. HGU(x) → HRB(x)
15
RESULTS 1: FORMALIZATION OF GWML2 IN FIRST-ORDER LOGIC (GWML2-FOL) WITH CLARIFIED AND REFINED SEMANTICS USING HYFO’S TERMINOLOGY A HydroVoid that is hosted by the Solid Body that is a submaterial of a Hydrogeo Unit or a Well. HGV(x) ≡ HV(x) ∧ ∃y,z [SB(y) ∧ hosts-v(y,x) ∧ submat(y,z) ∧ (HGU(z) ∨ W(z)]
AWB(x) ≡ WB(x) ∧ ∃y,v [A(y) ∧ ∀z [intra-const(z,y) ∧ WM(z) → submat(z,x)] ∧ HGV(v) ∧ P(r(x),r(v)]
Hydrogeo Void
contains
hosts-v Specialization of Matter
constituted-of
Aquifer Water Body
Aquifer
contains
EM(x) → M(x)
Earth Matter
Water Body that consists of all water in an Aquifer and resides in a Hydrogeo Unit.
Solid Body
A NAPO that is constituted of non-fluid Earth Material. SB(x) ≡ NAPO(x) ∧ ∃y[M(y) ∧ intra-const(y,x) ∧ ∀z (intra-const(z,x) → ¬FM(z))]
subclass-of
Hydrogeo Unit A specialization of Hydro Rock Body. HGU(x) → HRB(x)
16
RESULTS 2: LAYERING (MODULARIZATION) OF GWML2 CONCEPTS INTO ONES THAT ARE GROUNDWATER-SPECIFIC OR MORE GENERAL Groundwater module encapsulates GWML2’s unique focus and terminology but is also a (consistent) extension of HyFO for alignment with other hydro ontologies in the future [1] Foundational Concepts from DOLCE [2] Solid and Fluid (Geo-)Physics [3] HyFO Concepts
[4] Groundwater Specific Concepts
17
RESULTS 3: REVISIONS AND ADDITIONS TO GWML2 SEMANTICS
Introduction of several new classes to GWML2 at appropriate layer to separate groundwater-specific concepts from general water concepts Number of revisions to GWML2 terminology (disambiguation)
18
EXTENDING HYFO BY A NEW CONCEPT
HydroRockBody captures physical things that are partly solid, partly fluid, and partly space captures idea of a “water feature” (aquifer, river, lake)
19
SUMMARY OF RESULTS FROM USE CASE 1.
Formalization of GWML2 in first-order logic (GWML2-FOL) with clarified and refined semantics using HyFO’s terminology Logical extension of HyFO Available in Common Logic format from COLORE Tested consistency using theorem provers
2.
Layering (modularization) of GWML2 concepts into groundwaterspecific concepts and into more general ones Modules that encapsulates GWML2 unique focus and terminology but is consistent with HyFO and thus other hydro ontologies aligned to it in future Completion of the hierarchy
3.
Number of revisions to GWML2 Concept disambiguation Introduction of several new classes to GWML2 to separate groundwaterspecific concepts from general water concepts
4. Successfully tested HyFO for its suitability as a domain reference ontology for the water domain Proposed extending HyFO with a fifth core concept
20
TAKEAWAY, NEXT STEPS & OPEN QUESTIONS Domain Reference Ontologies are not just another standard Different purpose that domain ontologies: Analysis, refinement & integration of domain ontologies Not intended for direct use in applications
Apply same methodology to another hydro ontology to actually test/prove that the resulting ontologies are semantically interoperable: HY_Features? How transferable is it to other domains? Nothing really specific to the water domain Requires identifying well-scoped domains
General guidelines on developing domain reference ontologies? Requires broad skills and knowledge: formal logic, philosophical principles of upper ontologies and domain expertise Seem to be the hardest ones to develop: really require close collaboration between ontologists and domain experts There is no free lunch here
21
SELECTED REFERENCES [B15] Brodaric, B., 2015. GroundWaterML2 -- GW2IE Final Report. Tech. Rep. Open Geospatial Consortium Engineering Report 15-082, version 2.1. [BB12] Boisvert, E., Brodaric, B., 2012. GroundWater Markup Language (GWML) -- Enabling groundwater data interoperability in spatial data infrastructures. J. Hydroinform. 14 (1), 93-107. [BH14] Boyan Brodaric and Torsten Hahmann. “Towards a Foundational Hydro Ontology for Water Data Interoperability.” In: Proc. of the 11th Int. Conference on Hydroinformatics (HIC-2014). 2014. [HB12] Torsten Hahmann and Boyan Brodaric. “The void in hydro ontology”. In: Conf. on Formal Ontology in Inf. Systems (FOIS-12). IOS Press, 2012, pp. 45– 58. [HB13] Torsten Hahmann and Boyan Brodaric. “Kinds of full physical containment”. In: Conf. on Spatial Inf. Theory (COSIT-13). Springer, 2013. [HB14] Torsten Hahmann and Boyan Brodaric. “Voids and material constitution across physical granularities”. In: Conf. on Formal Ontology in Inf. Systems (FOIS-14). IOS Press, 2014. [HGB14] Torsten Hahmann, Boyan Brodaric, and Michael Grüninger. “Interdependence among material objects and voids”. In: Conf. on Formal Ontology in Inf. Systems (FOIS-14). IOS Press, 2014.
[HGB11] Torsten Hahmann and Michael Grüninger. “A naïve theory of dimension for qualitative spatial relations”. In: CommonSense at AAAI-SS 2011. AAAI Press, 2011. [KS03a] Yannis Kalfoglou and Marco Schorlemmer. “Ontology mapping: the state of the art”. In: The knowledge engineering review 18.01 (2003), pp. 1–31. [Noy04] Natalya F. Noy. “Semantic Integration: A Survey of Ontology-Based Approaches”. In: SIGMOD Record 33.4 (2004), pp. 65–70. [Tri05] A. Tripathi. “Developing a Modular Hydrogeology Ontology Extending the Sweet Ontologies. MSc Thesis, Georgia State University, 2005.
22