Upper Ontologies
2009/09/25 Leave a Comment
Ontologies are groups of concepts, relations and rules that define what we know. This can include definitions of types, or the known instances themselves. There are two main groups:
- Domain-Specific Ontologies (DSO) – Domain-specific ontologies define groups of concepts about a particular area of interest. These are defined by subject matter experts who use the definitions.
- Upper Ontologies (UO) – Upper ontologies try to define the universe of concepts at a very abstract level. Instead of describing concrete types, they concentrate on notions like Binary Relations and Temporal Entities. They are not directly useful for defining concrete concepts. Instead they are used to relate more concrete concepts by giving designers a set of high-level classifications for objects. The intent is that if they are used for a large number of DSO’s, they will enable rich inference across domains, allow data interchange and save everyone the effort of figuring out the basic concepts to use. UO’s are generally defined by philosophers and can have a heavy learning curve.
A list of some of the available UO’s is:
| Ontology | Language | Created | License | Concepts | Comments |
|---|---|---|---|---|---|
| SUMO | SUO-KIF | 2001 | Open Source | 1000 | Endorsed by IEEE |
| SUMO/MILO | SUO-KIF | 2001 | Open Source | 20000 | This is SUMO with a second expanded layer of concrete concepts. |
| Upper CYC | CycL | 1984 | Commercial | 6000 | In addition to the core, the full commercial system has hundreds of thousands of concepts. It is probably one of the most extensive in the world. |
| Open CYC | OWL | 2002 | Open Source | 6000 | The upper ontology should be the same as CYC, however it contains mainly just the concepts, not the rules that go with them. There is a lot of content that is not the upper ontology. |
| DOLCE | KIF | 2003 | Open Source | 100 | The upper ontology is small, but like CYC, is only one of the components of the full ontology. |
| GFO | FOL/KIF | 1999 | Open Source | 79 | Used in some medical projects. |
| PROTON | OWL | 2006 | Open Source | 300 | Research ontology |
| UMBEL | OWL | 2009 | open Source | 20,093 | Uses some components of OpenCYC as a basis. |
| SKOS | OWL | 2005 | Open Source | 32 | W3C ontology |
| COSMO | OWL | 2006 | Open Source | 5200 | See FTP site for an overview (COSMOoverview.doc) |
For further information on these ontologies, you can look at the web sites. Also, here are a few of the references I have used:
- Salim K. Semy, Mary K. Pulvermacher, Leo J. Obrst. (2004) Toward the Use of an Upper Ontology for U.S. Government and U.S. Military Domains: An Evaluation
- Viviana Mascardi1, Valentina Cordì1, Paolo Rosso2. (2006) A Comparison of Upper Ontologies
My recommendation? Choose one, it will save you time and extend the usefulness of your design. However, be prepared to commit some time to learn how the concepts work. If you are doing serious work with ontology in general, it helps to have a design strategy and a UO will give you that.
CYC is notable, not only because it is HUGE, but because it is comprehensive and has an excellent system of Micro-Theories which allow a federated approach to dealing with ambiguous and mismatched knowledge domains (fungus in medicine versus fungus in botany, for instance). However, it is bound to the proprietary CYC engine which makes interchange difficult.
One of the seriously important thing about CYC is that it has an excellent tutorial system. These documents cover a number of topics, including their use of events and time concepts. Anyone starting on ontology work should go through these materials.
I had to choose one a few years ago, so my choice was SUMO. I support it because it had been available for a while and is fairly comprehensive, but more because I am familliar with it at this point. I don’t think it is necessarily the best, but it is more than enough for the work that I do. I will expand on it in some of the discussions ahead.
