Thursday, March 16, 2006
The Second School of Semantic Science
ontologyMapping Glass Bead Games
Hello, (communication to the Protégé forum)
To a question:
>>
Hi,
>>
>>
which is the difference between a semantic network and an ontology?
>>
Thanks!
>>
>>
Kind regards,
>>
>> Yolanda
As pointed out by Paul, the names are used with many different meanings. Semantical networks historically are a graphical language in which a node represents a concept and an arrow (link) represents a relationship between the connected concepts. They have been given semantics but it has been proven (Woods 1975) that the given semantics were ambiguous. Different people interpreted the links differently, hence semantical networks defined at the end of the seventies were not usable for formal and clear knowledge representation.
From these concerns research went on and the representation logics have been refined and given more formal ground. One of the siblings are description logics, others are answer-set programming and first order logic with inductive definitions (the latter have more their roots in logic programming).
As graphical language a semantical network is very useful, but you have to give it a good formal account in order to do correct inference. This you can do by interpreting your graph as your preferred logic. But be aware, another may have chosen another interpretation hence you have to mention it.
Ontology is a similar ambiguous word.
The least interpretation is that it is a vocabulary of concepts, functions and relations for a given problem domain that you can use to represent particular information in that domain.
E.g. {father_of, mother_of, man, female, child_of} can be an ontology for representing human's families.
Typically such a vocabulary induces (not mentioned) formal relationships: e.g. Somebody who is a 'father_of' is a man. These implicit and natural relationships are often considered part of the ontology. Furthermore as for semantical networks, you have to state your 'logical context'. E.g. stating that you use an OWL-full DL ontology for a family domain precisely defines the boundaries of formal mathematics and the use (the problem domain). Leaving out OWL, full DL and your problem domain, may cause difficulties for other to interpret the truth ness of your information.
Eg. You probably assumed that on my proposed family domain the integrity constraint holds that somebody who is a father_of someone can NOT be a mother_of the same person.
Bert