Ontology of astronomical object types
Introduction
This page summarizes efforts on the construction of an ontology of astronomical object types.
The starting point is the Object Classification that is used in SIMBAD: a list of approximately 100 concepts
that are used as labels for individual astronomical objects in SIMBAD.
The ontology's main features are :
- it is an ontology of defined concepts (i.e. the concepts are defined by a set of necessary and sufficient conditions) which enables automatic reasoning
- it is implemented in the OWL 1.1 language (recent evolution of OWL-DL), but currently the complexity of the structure is still within the OWL-DL boundaries.
Concepts
The ontology contains a few top-level concepts which are:
- AstrObject
- subsumes all the concepts representing astronomical objects
- AstroPortion
- subsumes all the concepts representing portions of astronomical objects which are not astronomical objects themselves
- AtomicElement
- Atomic elements
- ClassificationCategory
- root concept for all kinds of classes used in objects desciption (e.g. spectral class)
- EMSpectrumRange
- sets of ranges in the electromagnetic spectrum
- Measurement
- measured observational parameters/properties
- Morphology
- geometry or morphology of astronomical objects (including non-astronomical object components)
- Process
- phenomenom or associated process
Additionally, a top-level concept called Test_classes can be found in some versions of the ontology file : it contains all the concepts used for tests (e.g. new concepts to classify to test the classification engines)
Properties
The following properties have been used in the ontology:
Object properties
| name | domain | range | inverse | transitive | description |
| hasEmissionIn | AstrObject | EMSpectrumRange | --- | no | The Astronomical object emits in the EM range specified |
| hasAbundanceOf | AstrObject | AtomicElement | --- | no | |
| hasMeasurement | AstrObject | Measurement | isMeasuredFor | no | Indicates that the DOMAIN AstrObject has the RANGE Measurement among its (observational) measurements |
| isMeasuredFor | Measurement | AstrObject | hasMeasurement | no | |
| hasProcess | AstrObject | Process | --- | no | Indicates that an astronomical object is the location of a Process/has an attached process |
| hasPortion | AstrObject OR AstroPortion | AstroPortion | isPortionOf | no | |
| isPortionOf | AstroPortion | AstrObject OR AstroPortion | hasPortion | no | |
| hasMorphology | AstrObject | Morphology | isMorphologyOf | no | |
| isMorphologyOf | Morphology | AstrObject | hasMorphology | no | |
| hasComponent | CompoundObject OR AstroPortion | AstrObject | isComponentOf | yes | Indicates that the DOMAIN CompoundObject has the RANGE AstrObject among its physical components (the RANGE AstrObject cannot be a component of a component of the compound object) |
| isComponentOf | AstrObject | CompoundObject OR AstroPortion | hasComponent | yes | Indicates that the DOMAIN AstrObject is a physical component of the RANGE CompoundObject (not a component of a component of the RANGE CompoundObject) |
| hasSpectralLine | AstrObject | AtomicElement | --- | no | |
| hasEmissionSpectralLine | AstrObject | AtomicElement | --- | no | |
| hasAbsorptionSpectralLine | AstrObject | AtomicElement | --- | no | |
| isIonizedBy | InterStellarMedium | EarlyTypeStar OR Shockwave OR PlanetaryNebulaCentralStar | --- | no | |
| hasClassification | AstrObject | ClassificationCategory | isClassifiedAs | no | |
| isClassifiedAs | ClassificationCategory | AstrObject | hasClassification | no | |
| hasProgenitor | Supernova | WhiteDwarf OR WolfRayetStar OR HighMassStar | --- | no | |
Datatype properties
None
Annotation properties
Some annotation properties are used to store references to concepts in existing applicationt
| name | description |
| adc:keyword | Astronomical Data Center keyword |
| gcvs:code | General Catalogue of Variable Stars code corresponding to a concept of the ontology |
| interface:naturalName | name in natural language for display purposes |
| interface:searchKeywords | keywords for search purposes |
| interface:unit | the unit associated with the value of an instance of measturement |
| interface:values | The operators and values for an instance (of measurement) |
| misc:description | textual description/definition of a concept |
| misc:externalTests | tests that should be taken care of outside of the ontology (e.g. check that a value is within sensible boundaries) |
| ned:code | object type code in NED |
| simbad:name | Standard name in SIMBAD's object classification |
| simbad:shortCode | Short code in SIMBAD's object classification |
| simbad:luminosityClass | Main part of the luminosity class in SIMBAD (letters only, found within SIMBAD's spectral type value after said spectral type) |
| simbad:morphologicalType | morphological type for galaxies in SIMBAD |
| simbad:spectralType | Main part of the spectral type in SIMBAD (letters only, found at the beginning of SIMBAD's spectral type value) |
| vizier:keyword | VizieR registry keyword |
Technical issues
Choices and lessons learned during the construction of the ontology.
Protégé issues and advice
The Protégé editor is a great tool for building ontologies but some
ProtegeAdvice helps to face a few potential problems
Note: the
DLSyntaxDisplay option the OWL>preferences menu was bugged for a lot of revisions until ver. 3.3 and may even lead to data loss! So if you want a DL-like syntax representation of restrictions on properties, select the CompactOWLClassDisplay instead. Else, keep the
ManchesterDisplay, it is the safest of all.
OWL flavours
The Web Ontology Language is availiable in 3 (soon 4) flavours based on differents
DescriptionLogics :
| flavour | description logic | decidable | comment |
| OWL-Lite | SHIF(D) | yes | least expressiveness of the OWL flavours, least resource-consuming |
| OWL-DL | SHOIN(D) | yes | a lot more expressive, more resource-consuming but still decidable |
| OWL-1.1 | SHROIQ(D) | yes | revision of OWL-DL, adds qualified cardinality restrictions and more expressivity on roles, even heavier resource-wise but decidable |
| OWL-Full | beyond SHROIQ(D) | no | No limit on expressiveness, only subsets are decidable |
Various inference engines
We have performed various
InferenceEngineTests.
The most efficient reasoners available for description logics are the following. All are based on different description logics and implementations.
| reasoner | test version | logic | implementation | license | comments |
| RACER | 1.7.24 | SHRIQ(D) | CommonLISP | free license | discontinued since 1.7.24 (authors went commercial with RacerPro) |
| RacerPro | 1.9 | SHRIQ(D) | LISP | commercial | DIG-only interface doesn't need a commercial license |
| FaCT++ | 1.1.3 | SHOIQ(D) | C++ | GPL | difficulties with large-scale hierarchies |
| Pellet | 1.3 | SHOIN(D) | Java | M.I.T. (free) | Theoretically now supports SHOIQ(D) but Qualified Cardinality is not really supported yet, difficulties with large-scale hierarchies |
| Pellet | 1.5.2 | SHOIQ(D) | Java | M.I.T. (free) | Even more powerful than RACER 1.7.24, good support, implements configuration flags to simulate assumptions such as closed world or unique name |
The tests compared the performances of the different reasoners for the following tasks :
- checking the consistency of the ontology
- classifying the ontology
Conclusions:
- (Given the expressiveness we need RACER/RacerPro was the best option until the release of Pellet 1.4. RacerPro being commercial, prices and possible incompatibility with some API (e.g. Jena) might be a problem. Basically, the best option is RACER 1.7.24 since it's free and, though discontinued, has both good performance and extensive compatibility with the OWL APIs) *Currently, Pellet seems much more interesting than RACER since it has basically the same level of DL handling, similar if not better performance and obviously better support
- To limit complexity, existential restrictions (\exist R C) are to be used instead of qualified cardinality (\geq n R C) or (\leq n R C) whenever possible.
Download
The following versions of the ontology can be dowloaded for testing.
Most recent files are on top of the table.
Applications
Details of
OntologyUseCases for the ontology of astronomical object types are available on the
OntologyOfObjectTypesUseCases page.
Links
--
AlexandreRichard - 01 Jun 2006