Two modern approaches to semantic modeling - ORM and ERMM

Semantic modeling occurrence is obliged to P. Chen and his model «Entity - Relationship». He laid the bases for a semantic technique of the DB schemes designing in his initial article together with the representation of a new model. This technique excellent supplements a classical technique of normalization. Two factors determine a success of semantic technique application (besides designer knowledge and skills): used semantic model expressiveness and minuteness of the semantic scheme analysis in used rules of its translation in the DBMS model. For this reason, synthesis of the maximum expressive semantic model is of interest. Thus, fixing in the scheme of all basic concepts and all laws of mutual relations between them will be possible within the framework of this model at a stage of initial data domain formalization. The ideal data domain semantic scheme should contain the full formal data domain description, so that all subsequent intermodeling transformations of the scheme had be only syntactic and did not demand the repeated analysis of data domain semantics and works with texts in a natural language, in which the revealed business rules are preliminary fixed. The opportunity of construction of such semantic schemes will open prospect of full automation of DB designing process, including generation of such realization elements as views, triggers, etc. Using Date's pattern for synthesis of semantic model as a technique of the analysis, the author compares two modern approaches to semantic modeling: Halpin's model «Object - Role» (OR-model) and own model «Entity - Relationship - Mapping» (ERM-model). Both these models are distinguished from previous semantic models with the richer set of the structural and constraining concepts, allowing to present in the data scheme practically all semantics of the modeled world. In the ERM-model concepts «object» and «mapping» provide all expressive force of language, and other formal concepts are derivatives from them. For this potential they are obliged to functional treatment of natural language semantic categories. Semantic concepts of OR-model «object» and «role» are not enough for representation of the most part of the information. For this purpose, Halpin already at a stage of model formal concepts definition entered the concept «predicate». But thus the model loses the individuality and originality, coming nearer on the base concepts to the ER-model and relational model. Understanding the importance of various objects orderings in predicates, Halpin enters them in the model as the perusals of predicates. But he does not go beyond usage of these perusals for verbalization of the scheme and the data. He does not see mappings underlying it, and they open additional opportunities for the formalization of business rules. Distinctive feature of Halpin's model integrity constraints is, that he is invented the special term and graphic primitive for any useful feature of the data. That leads to the unhandiness of the model definition and complexity of its application. As against the OR-model and ERM-model, we offer the limited universal set of constructions. The basis of ERM-model composes the generalized concepts of logic and mathematics, combined in use, as well as in a natural language, provides all its expressive power. If Halpin does not consider data operation at all, in ERM-model they play the important role even at the construction of the scheme as with their help derivative mappings are defined, and they are the object of integrity constraints as well as the base mappings. Understanding the importance of powerful semantic model synthesis and considering as the standard of expressiveness a natural language, authors of the ORM and ERMM go by various ways. If Halpin decomposes the predicate (or relationship) of degree n on the set of n roles, which extensions aggregation does not restore predicate extension, in the ERMM instead of the relationship type of degree n is considered as 2 -2 types of mappings, which each copy is equivalent to the whole relationship in information sense. Instead of n roles we have 2 -2 mappings which characteristics more full describe semantics. If to this, the opportunity of determining the derivatives of mappings, using algebra maps, is added, then descriptive abilities of ERM-model become practically boundless. The carried out analysis shows that the opportunities of data structures and integrity constraints definition of the ERM-model are not worse, than ones of the OR-model, and sometimes even surpass them.

Keywords

Authors

References