Measuring Semantic Gap: An Information Quantity Perspective

There is much interest in semantic gap shown in the literature, and yet little seems to have been done on measuring it with a scientific means. A conceptual data schema specifies the structure of the 'data base' for a computer based information system and provides abstraction of some part of the real world. Data constructs, which we call 'data level connections' ('data connections' for short), in a data schema are used to refer to objects and the relationships between them in the real world, which we term 'real world relations'. That is, the latter is taken as the semantics of the former. This 'referencing and referenced' relationship may not be perfect, which may be seen as a kind of semantic gap. We suggest looking at information systems with an information theoretic framework, i.e., information creation, and transmission through representation. We define a semantic gap in an information system as the inefficacy in information transmission through representation, especially when an information bearer (carrier) is unable to carry the information that it is supposed to. We investigate how such a situation could happen, and identify the condition regarding information quantity for a semantic gap to be zero between a 'data connection' and a 'real world relation'. Based upon this, we give a measure for semantic gaps. These constitute an information quantity perspective for looking at including measuring a semantic gap.