Ontologies

It is very difficult to describe a set of web pages using a semantic web language. First the language has to specify a vocabulary and then formally define the vocabulary so that it can be used by intelligent web agents and applications. This vocabulary is what is commonly referred to as ontology. For each topic that a web site can be based, an appropriate ontology should be selected. This is the ontology domain. From this domain, selections as to descriptive adjectives are made and semantic markup is created.

Semantic languages vary in expressiveness, semantic richness and complexity. Obrst (2003) created the Ontology Spectrum which is a framework for depicting various information models in terms of increasing semantic richness. A modified version of this Ontology Spectrum is shown.

Following the spectrum for Weak Semantics to Strong Semantics, semantic richness increases. Weak semantics indicate the expression of very simple meaning while strong semantics indicate the expression of arbitrarily complex meaning (Obrst, 2003). As you progress up the spectrum, the domain semantics are based on logic, allowing the machine to make valid inferences and execute sound semantic constraints (Obrst).

Semantic Normalization

Normalization is the area of the semantic web that discusses ontologies created and managed by separate and distinct sources and how they can be integrated with each other. Different web designers create web pages on the same topic. How will the knowledgebase know that the two ontologies developed are referring to the same subject domain? Metadata normalization will determine that two ontologies from separate sources are referring to the same subject domain. Consider the case where two different web pages refer to the same person but in different contexts. The ontologies used may be quite different and the knowledgebase needs to be able to determine which ontology, or combinations of ontologies are applicable. Take for example the name “Benjamin Franklin”. The knowledgebase needs to be able to differentiate between the subject matter of Benjamin Franklin the politician, Benjamin Franklin the Postmaster General, Benjamin Franklin the scientist and Benjamin Franklin the husband, father and friend.

A different type of metadata normalization is the case where there are multiple names of a single entity. The knowledgebase needs the ability to recognize that the list of names should be attributed to one widely known name. An example is Michael Jordan. The names “Your Airness”, “Air Jordan”, the phrase “I want to be like Mike” are all referring to Michael Jordan.

Ensuring interoperability between independently developed ontologies is called semantic normalization. Ontology, provided by one designer, can be compared and integrated into an ontology provided by the knowledgebase, or if one does not exist, can be added to the knowledgebase. As more semantic web pages are examined, the list of ontologies will grow and as similar web pages on like topics are examined, the ontology for that topic will be refined. As semantic web languages evolve, there likely will be many public libraries of ontologies available.

Heflin and Hendler, 2001, in A Portrait of the Semantic Web in Action believe that hierarchies will develop among ontologies. Once that happens, any two ontologies with common ancestors will automatically be interoperable.

References

Heflin, J., & Hendler, J. (2001). A Portrait of the Semantic Web in Action. IEEE Intelligent Systems, , 54-59.

Obrst, L. (2003). Ontologies for Semantically Interoperable Systems. Proceedings of the Twelfth International Conference on Information and Knowledge Management, November 2003, 366-369.