The need for formal methods for Object Oriented (OO) systems resulted in methods like UML and Lepus3 that are de-facto graphical languages equipped with formal tools that are able to handle the design of OO systems. However, they lack precise semantics which might lead to problems, such as inconsistencies or redundancies. On the other hand, to our knowledge, there is no approach that allows one to understand and follow the requirements of a design-implementation path for people that lack knowledge about standard software modeling language. The approach to OO system modeling presented in this paper uses Controlled English (CE) (a well defined subset of English) in the area where graphical languages are currently used. Object Oriented Constructs are modeled first in Description Logic (DL) that provides the logical framework and the CE verbalization of DL (CEDL) finally bridges DL with CL allowing to access OO world in formal, yet understandable way for both human and computer.
Rapid development of software engineering and software production
methodologies, which took place as complexity of information systems advanced, is
connected with the need for formal description methods for knowledge acquired
during their development [
The approach to OO system modeling presented in this paper differs from the
above approaches as it uses the Controlled English1 [
Software structures (especially OO) can be treated as ontologies3. Description
Logic (DL) is a subset of first order logic focused on ontology formalization and has
the very important property of being computable. Computability ensures that
reasoning tasks4 of DL can be made in finite time and space. Some dialects of DL
(e.g. EL++ [
Although DL is now most often associated with the semantic-web, new
applications appeared recently proving the usability of DL also in other fields of
2 Bioware is a neologism for the human that interacts with the software-intensive system.
3 The source-code of a computer program is a set of sentences describing what to do with data
and therefore they form an ontology of program behavior and data.
4 Reasoning tasks include for example.: concept classification, that is, a hierarchical
arrangement of concepts within the notion of includes. Another one is classification of
instances to certain concepts.
5 Responsiveness – the ability of a computer system to perform an assigned function within the
required time interval
6 DL is equipped with open-world assumption and it lacks defaults while OO uses closed-world
assumption and uses defaults to describe a class-inheritance.
interest. These applications include DL verbalizations in CE (CEDL) that enable access
to DL in natural language as a part of human-computer interface. Very expressive CE
like ACE [
The research for CEDL described in this paper was inspired by ACEOWL, however
the grammar of CE used in this research (called LL(1)CEDL), was implemented using
LL(1) top-down parser generator [
Object (the main idea used in OO languages) and class (specification for object construction) together with carefully selected relations (like extend or materialize) are modeled here first in DL and then the adequate LL(1)CEDL sentence is created. The existence of a bidirectional LL(1)CEDL<->DL mapper, that is a part of a parser, bridges these two technologies together.
If for each class, object, method and attribute a corresponding instance of concept is assigned then the basic model of the object can be created in DL (see Fig.1). Next, the relations between classes can be established. Classes can extend8 each other to provide a subtyping mechanism9. On the other hand objects realize every type related to their class and therefore adequate properties on selected relations are required. Using selected relationships it is possible to specify requirements for a class. E.g.: to specify that a class is abstract10, to require that the class must be a root of a class hierarchy or to prevent the class from being inherited11. One can also require that a class is an implementation of a singleton design pattern, explicitly requiring that it can have one and only one unique realization. It is also possible to handle sets of classes e.g.: one can compute the set of classes that form a branch of the class hierarchy and 7 Additional features include: ―A is equivalent to B‖ construction that correspond to AB DL expression, ―the C‖ that corresponds to unnamed individual of concept C and allow use of parentheses that allows production of more complex expressions in CE. 8 I use the term ―extend'‖ in place of ―subclass‖ because it has more intuitive meaning about the source and the target than a subclass relationship (―B subclasses A‖ is not so intuitive as ―B extends A‖) and is easier to understand for a non-expert. 9 Subtyping mechanism (partial ordering over types) is a core idea behind the paradigm of a
OO abstraction. 10 A class that cannot have any instances. 11 Such classes are called ―final‖ in Java. hierarchy itself. Having the ability to express relations between class hierarchies, the opportunity to describe structural relationships of design-pattern is opening. It is even possible to set up the architectural design-constrains like layer-separation that then is able to be computed by DL Reasoner.
Every object that materialize class C have object that fills attribute had by C and
identified by P.
Everything that fills an attribute identified by P and had by C is an object that realizes X.
Everything that is identified by P and had by C
is equivalent to one attribute. identify.{P}have.{C}{[attribute]} fill.(attributehave.{C}identify.{P})objectrealize.{X} objectmaterialize.(class{C})have.(objectfill.(attributehave.{C}identify.{P})) C is a class.
{C}class P, Q, M, N are signatures.
{P,Q,M,N}signature
C -P:X -Q:Y +M() +N()
Everything (that implements a method identified by M and had by C) that is had by an object that materializes C is equivalent to one function.
Everything that is identified
by M and had by C is equivalent to one method. identify.{M}have.{C}{[method]}
This paper presents the results of research on the potential application of Description Logic verbalized by Controlled English inspired by ACEOWL and implemented with LL(1) grammar (called LL(1)CEDL) as a modeling language for Object Oriented (OO) systems. It was shown that LL(1)CEDL can be used to describe basic OO constructs. The research made so far gives hope that CE in general enables us to use one and the same system for storing requirements, as well as the project and system architecture, which in turn ensures the logical cohesion of artifacts created in different stages of software development. Moreover, in such a case, it would be possible to maintain a coherent terminological base between different groups of people involved in the information project through a common knowledge management system. These possibilities are currently being intensively investigated by the author of this article. To compare the usage of CE as an OO modeling language in comparison to existing methodologies that use the graphical languages (UML/LePUS3) the evaluation is also planned.