This work is related to the creation of a new technology for the development of integrated expert systems (IES) based on the further development of the problem-oriented methodology and intelligent software environment of the AT-TECHNOLOGY workbench through the integration of an ontological approach to software design for applied IES and methods of intelligent planning and management of IES development processes. with different architectural typology. The description of prototyping processes for IES based on the use of the basic components of the model of an intelligent software environment with an emphasis on expanding one of the components of the technological knowledge base by developing an applied ontology of typical IES architectures and implementing interaction with an intelligent scheduler is presented.
Methods of intelligent planning, and their integration with knowledge engineering,
proposed and described in detail in [
To build an IES, a problem-oriented methodology has been created, actively used
and is constantly developing [
A significant place within the problem-oriented methodology is given to methods
and means of intellectual support for the most labor-intensive stages of the life cycle
of building an IES - analysis of system requirements and design. Here, the concept of
an intelligent environment model [
The accumulation of experience associated with the development of various IES architectures for specific problem areas and classes of problems to be solved has shown that the processes of prototyping IES have the greatest complexity due to the high laboriousness of building an IES architecture model performed by a knowledge engineer, as well as the need to use a large number of individual software and information components of the AT-TECHNOLOGY workbench, which implement the basic functionality of the IES of various architectural typologies (static, dynamic, training, etc.).
It is also important to note that the complication of management processes for software development of applied IES at certain stages of the life cycle depends significantly on the sources of knowledge used and the integration of various technologies for the automated acquisition (identification) of knowledge from experts, natural language texts (Text Mining technologies), databases (technologies Data Mining, Deep Data Mining, etc.). Since, as a rule, these technologies arose and developed independently of each other, such autonomy and distribution significantly complicates both sharing, leading to an increase in labor intensity in the development, maintenance and monitoring of such voluminous resources as knowledge bases (KB), DB, etc. software tools, including their repeated component-wise use [4].
Therefore, the problems of integrating the methods and technologies of Text
Mining, Data Mining, etc., as well as research in the field of creating tools and
technologies for distributed knowledge acquisition are most relevant today, as
evidenced by a number of works, for example, [
A similar problem of integrating models, methods and software arises in the design
of software for dynamic IES [
One of the new approaches to the creation of intelligent technologies for the
development of IES is the further development of the problem-oriented methodology
and the intelligent software environment of the AT-TECHNOLOGY workbench
based on the expansion of the technological knowledge base [
Proceeding from this, the main emphasis in this work is made on the discussion of issues related to the experimental software research of prototyping processes for IES based on the integration of the ontological approach and methods of intelligent planning and management [3].
It should be noted that in modern research considerable attention has begun to be
paid to both the creation of ontological models of software design processes in
general, for example, [
Currently, a new direction is emerging, called
Ontology-Based-SoftwareEngineering (OBSE) or Ontology-Driven-Software-E Engineering (ODSE) [
As a rule, all semantic models of the processes of creating software systems and /
or their components are based on a specific classification of the used ontology
models, including specific methods for implementing ontological models of design
processes, for example [
Let us briefly consider some of the important features of the technology for
supporting the prototyping processes of applied IES using the main components of the
intelligent software environment [
The basic declarative component of the model of an intelligent software
environment in accordance with [
The initial data for generating IES prototype development plans are the IES
prototype architecture model, described using the hierarchy of extended data flow
diagrams (EDFD [
The implementation of the tasks of the plan is carried out using a set of operational
(instrumental) RUC. Operating the SDP as the main algorithmic element, the
intelligent planner at each moment of time makes a detailed construction of the IES
development plan, depending on the current state of the project (the type of the NF
problem being solved [
The general architecture of the AT-TECHNOLOGY workbench is built in such a
way that all functionality is distributed, i.e. “Spreads” to components registered in the
workbench environment and operating under the control of an intelligent development
support environment. At present, within the framework of the intelligent technology
for constructing an IES, two groups of RUCs are used - components that implement
the capabilities of the procedural RUC, and components that implement the
capabilities of the information RUC [
All SDPs are classified as follows [
The increasing complexity of IES architectures, and the appearance of a large
number of RDPs and RUCs in the technological knowledge base led to an increase in
the complexity of the search, and therefore the methods and algorithms for planning
[3] used by the intelligent scheduler were improved by implementing a fairly
wellknown approach related state space [
The essence of the method for generating an action plan for a knowledge engineer is to perform a sequence of transformations of the architecture model of the current IES prototype, performed in 4 stages [3]: obtaining a generalized EDFD in the form of a graph; generating an exact coverage (i.e., a set of RDP instances with mutually disjoint fragments containing all the graph vertices) using heuristic search; generating a knowledge engineer plan based on the resulting detailed coverage; generating a plan view.
All of the above steps are performed by an intelligent scheduler that fully implements the functionality associated with planning IES prototyping processes. With the help of the preprocessor of the EDFD hierarchy, the preprocessing of the EDFD hierarchy is performed by transforming it into one generalized diagram of maximum detail. The task of covering the detailed EDFD with the existing RDP is implemented using the global plan generator, which, based on the technological knowledge base and the constructed generalized EDFD, ensures the fulfillment of the task, as a result of which an accurate coverage is built, which is subsequently converted into a global development plan.
The detailed plan generator provides detailing of each element of the coverage, i.e. on the basis of the obtained EDFD coverage and technological knowledge base, each element of the coating is detailed, thereby forming a preliminary detailed plan.
Then, based on the analysis of available RUCs and their versions (data on which are requested from the development process control component), the plan interpretation component forms a detailed plan, where each task is associated with a specific RUC and can be performed by a knowledge engineer. With the help of the component for constructing the final plan, the necessary representation of the plan is formed for its use by other components of the intelligent software environment (component of the plan visualization, etc.).
As noted above, the complication of the architectures of the designed IES and the appearance of a large number of RUCs in the technological knowledge base led to an increase in the complexity of the search and increased the negative effect of the nonoptimal choice of solutions. Therefore, the composition of the technological knowledge base was expanded by building an applied ontology of typical IES architectures [2].
A detailed description of the methods and algorithms for the implementation of intelligent planning tools is given in [2,3] and others, and below, in the context of this work, some results of experimental modeling based on the implementation of the ontology of typical IES architectures are considered. 3.
As a basic model of applied ontology, we took a model developed within the
framework of the problem-oriented methodology for constructing IES [
The current version of the applied ontology [
The modified model is a semantic network described in the form: Mom = <Vom, Uom, PDom>, where Vom is a set of elements of an architecture model (Mies), built on the basis of the ideas of deep integration of components (at all levels of integration), and each element includes name of the ontology vertex, weight (in the range 0 ... 100) and information about the RUCs used; Uom - a set of links of several types between elements of the Mies model (parent and child nodes of the ontology), and the semantics of the types of these links can vary widely (aggregation, association, hierarchy, strong, medium and weak links, etc.) depending on the RDP used; PDom (optional) a lot of special data, i.e. information of a different nature, specifying features and / or non-standard approaches to the development of individual components of the IES prototype (parameters, texts, information about external subsystems, components, applications, etc.).
To build an applied ontology on the basis of this model, tools have been developed to support the construction of an applied ontology of typical IES architectures. 4.
As already described earlier, the essence of the problem is to implement an approach
to the construction of applied ontologies based on taking into account the features of
the architecture model of the designed IES and the features of the currently
implemented component-wise functionality in the form of a set of RUCs. The
architecture model of the IES prototype is presented in the form of the EDFD
hierarchy [
The peculiarity of the IES architecture model lies in the multi-level integration, manifested in the EDFD hierarchy and, as a consequence, the identification of architecture elements at different nesting levels, which leads to different architectural solutions, including the use of RUC. Accordingly, the same elements of the architecture can have different functionality; therefore, RUCs that implement this functionality for identical elements also work in different ways.
Therefore, the structure of the applied ontology of typical IES architectures, as well as the algorithms and procedures for its creation and storage, should be developed so that, through their use, it is possible to configure the RUC depending on the structure of the IES architecture model.
On the other hand, the advantage of the basic ontology model lies in the fact that
instead of the general vocabulary of concepts used in most information processing
systems of the ontological type, a semantic network is used, which made it possible to
significantly strengthen the semantics of vertices and display a significant number of
relations not only of the taxonomic type, as in most ontologies, but also relations
reflecting any declared specificity, as well as using powerful functionality in the
interpretation of relations and nodes (elements of the ontology [
There are three types of links between the elements of the ontology: a link of the "part-whole" type (aggregation), for linking the elements of the ontology that are at different but adjacent levels; a link of the "association" type, for linking ontology elements at the same level; "Weak" link, for linking elements that are both at adjacent levels and at the same level.
In addition to the previously described types of links in the ontology of typical IES architectures, interlevel links between ontology elements located at different levels will be implemented.
Now let's look briefly at some of the implementation features. Since the ontology
of typical architectures is part of the technological knowledge base, it has access to
data on the RDP and RUC. The technological knowledge base is stored in the form of
an XML document [
Since the ontology of typical architectures contains various models of IES architectures and their elements (subsystems / tools / components), the following operations can be performed on elements of architecture models, as on elements of an ontology: initialization of adding a new architecture to the ontology; adding architectural elements to the ontology; removal of architectural elements from the ontology; sampling of ontology elements; unification of ontologies. 5.
Thus, the experimental base in the form of accumulated information and software for applied IES, developed on the basis of the problem-oriented methodology and the AT-TECHNOLOGY workbench supporting it, turned out to be a successful "testing ground" for the continuation and development of research in the development of elements of a new intelligent planning and control technology. processes of building intelligent systems, including those based on the ontological approach.
In fact, a technological transition was made from "automation" to "intellectualization" of labor-intensive design processes and maintenance of information and software for applied IES, by creating conditions for the effective use of an intelligent scheduler, in particular, to create a technological knowledge base (RDP, RUC information and operational character, ontology of typical architectures), and then carry out full-featured research to create elements of a new technology. 6.
This work was supported by the Russian Foundation for Basic Research (project no. 18-01-0457). 2. 3. 4. 5. 6.