Ontology Learning in Practice: Using Semantics for Knowledge Grounding

Background

Ontology-based knowledge grounding avoids problems with concept ambiguity and lack of semantic relations between concepts. The use of ontology enables the unification of different representations for the same object and allows easy data exchange between different systems. In this case, ontology can be interpreted as a concept database for formal development of information, preferences and knowledge (IPK) (Gadomski, 1999). The main idea of cognitive IPK architecture is the generalization of available data by applying information and preference relations for the choice of proper statements. Information categorization and prioritization is used in order to get explicit reasonable statements for a particular domain. In the Top-down Object-based Goal-oriented Approach (TOGA meta-theory), ontology is considered to be a goal-oriented and role-depended set of concepts (Gadomski, 1993). The TOGA descriptive knowledge is defined by relation rules, physical laws, theories and models, and its operational knowledge is expressed by algorithms, methods, instructions, procedures and actions.

The ISO 1087-1 standard (2000) defines a concept as a “unit of knowledge created by a unique combination of characteristics” (p. 2). It is language independent, but can be influenced by the context. Thus, the ontology learning process can be described by the following steps: a) selecting the knowledge object, b) learning information about this object – defining concepts, terms, and a set of constraints, c) defining rational criteria for the evaluation of learned knowledge d) documenting (Ushold & King, 1995).