MLM Summary

MLM Summary

We have shown that PowerType Absorbance (PTA) is a powerful mechanism to significanty reduce conceptual models. The use of Partitioning Classes (PC) has the same effect. Layer mistakes, as they are pre-programmed with other more complex MLM approaches, are simply unlikely to appear with our approach, since all classes are lying in the Schema Layer (SL) and all IIs are represented on the Particulars Layer (PL).

With the notations introduced it is possible to have UML-like graph visualizations. Compared to UML, our advantage is that information of the SL and IL can be represented in the same diagram. UML normally requires two different diagrams for that, where a class from a class diagram cannot be connected to the IIs of the object diagram.

Leanness and Performance: For those who have to handle ontologies on a very large scale, such as goods, species, diseases, medicines, etc., storage and performance requirements are an important issue. If one wants to maintain an ontology of several million species, traditionally a meta class is required for each species to store the number of individuals, and for each of the several million classes, the OP instance to connect the class and its meta class must also be stored. Within our approach, the additional metaclasses and OP instances can be saved. Since the information about the number of individuals (.^Individuals) is stored in the same class instance, we can access it with a single query without using the OP ◊Instance. This generally results in a significant reduction of memory space and better performance for queries. This has been extensively used in practice for handling 100 millions of triples, even under realtime requirements in the domain of Natural Language Generation (NLG) for the generation of stock reports and weather reports [Bens2017]. Under interoperability aspects, experience has shown that an arbitrary number of ontologies from different website could be accessed at the same time, with scalable performance in cloud environments.

In this section we have proposed a Data Property (DP)-centric approach to MLM. Data Properties have been generalized into the three-faceted properties PDP, TDP and CDP. We have shown that this approach makes ontologies less redundant, more transparent, and easier to understand and to edit. The approach eleminates the need to use PowerType Absorbance (PTA) and modeling patterns like materialization [PiZi1994], and subsumes the aspect of Dual Facet Behaviors (2FB) of classes/instances. The validity of a conceptual model can easily be checked by the syntactic names of KEs. Of course, it is the responsibility of the modeler to choose proper names that reflect the intended semantics. We think that our modeling guidelines and naming guidelines are easy to learn. Applying the simple transformation rules and guidelines, it is even possible to automatically transform models from our notation to models designed with traditional methods, and vice versa. By applying the methods and guidelines proposed, modelers obtain significantly reduced ontologies, preserving semantics.