New generation of ES

The concept of New Generation Expert Systems is a priority development, which has no analogues in the world. It organically combines such leading areas as neurocomputing, DBMS and computer graphics . The basis of their own development, the results of which successfully defended the dissertation.

Among other things, based on this technology, a large effect of information compression is achieved, incl. and in real time, which is very important when transferring data. I'm not talking about the movement of audio, video files and VoIP (Voice over Internet Protocol, or IP-telephony). On the basis of this technology, it is possible to carry out intelligent public opinion polls, which will significantly reduce the cost of processing the survey results in comparison with the traditional method of installing meters in households. And much more. Despite the marked diversity of areas of application of the new technology, this page describes the concept of creating on this basis only ES NP.

Terminology

• Artificial Intelligence - AI
• Information Technology - IT
• Expert System - ES
• Database Management System - DBMS
• Database - DB
• Knowledge Base - BR
• Subject area - software
• Workstation - AWP
• Expert System of the New Generation - (so far conditionally) ES NP

Basic properties

What is the difference between a new approach to building an ES from a traditional one? I present only the main advantages of the proposed approach:

• With one of the global views, the majority of ES can be divided into static and real-time ES. This development can be used in both cases.
• Building a software model organically combines both traditional approaches to building such models: attributive and structural (see, for example, E.V. Popov).
• Setup for various software. DB and BR can be formed directly for the desired software, and can be formed for several of them at once.
• Automatic formation of a software concept system. Unlike most methods of extracting knowledge from experts, in which the latter have to squeeze something out of themselves, with the proposed approach, the system of software concepts is formed automatically, including and dynamically.
• Formation of semantic links between software concepts. This procedure also does not require the participation of experts. Semantic links between concepts, like the concepts themselves, are formed automatically and can change dynamically. Link changes also occur automatically.
• The possibility of modeling the "outside world" (model of the open outside world). There is an opportunity to work both with the model of the "open external world" and with the model of the "closed external world". The technology of automatic database prediction is built into the technology of the formation of the database and knowledge base. Moreover, the forecast is carried out in the formed concepts. Forecasting in this case is significantly different from only recognizing the situation that occurred in the vulgarity. It is possible to predict both partially new and completely new situations.
• Cognitive abilities. There is an automatic structuring of software data. In addition, each concept is formed in conjunction with the parameters on which it depends, and which are also formed automatically. Taking into account the ability of automatic prediction, automatic recognition of concepts takes place according to the available parameters.
• The independence of the concepts formed from the subjectivity of experts and knowledge engineers. The concepts are generated automatically, based on the available objective information about the software.
• Intermediate is eliminated when creating a database and knowledge bases: knowledge engineer, because the formation of data bases and knowledge bases is automatic.
• Real associativity. DB and KB are formed according to the associative principle. Implemented a fully associative principle of information retrieval.
• Due to the associative principle embodied in the very structure of the database and knowledge base, as well as the associative method of searching for information, there is no need to implement numerous search strategies and decision rules: The principle of formation and the principle of search are the same.
• Toolkit for AWP. To create an automated workplace, a specialized toolkit is required, which is also a graphical interface to the automatically created KB and DB. You can look at it as a kind of graphical tool for "digging into your own thoughts."
• Flexibility to customize as an expert. (this is already a property of the toolkit for an automated workplace). The customization of an automated workplace for an expert is made by him using the graphical interface in an interactive mode. The setup made by an expert is that he himself links his understanding of the software with the one offered by the system described here. The attachment of concepts is performed using the "point and execute" method (this is what is meant by the graphical interface). Using the same toolkit, another expert, also independently, links his understanding of software with DB and BR. Thus, each expert creates his own software presentation for himself. At the same time, the very meaning of the word expert expands. An expert is now understood not only as an expert involved in the formation of a database and knowledge base, but also as a direct software researcher.
• Compression Information. As a result of automatic elimination of redundancy, a large degree of information compression is achieved in the database and in the database. The technology of building a KB and a DB is such that, starting from a certain amount of data, the growth of the volume of KB and DB ceases to depend linearly on the amount of input information.
• What can serve as a description of the software. As a software description, any full-text databases related to this software, any ??? texts written on the topic related to this software. A separate description of a particular software can also be a large array of numerical data, if the software is a process.

Limits of applicability

• It requires a sufficiently large amount of information related to specific software.
• The software must contain internal laws that must be present in the available or hidden form in the existing descriptions of this software.

Areas of use. It is difficult to structure and poorly formalized software with a large amount of information. Some of them were listed in the announcement.

Other uses:

• automated formation of a system of concepts and their signs (parameters), and on their basis various recognition and prediction systems. (It can be used, for example, in the stock market, in currency dealing, for text and speech recognition, and in all the areas listed above and below, since it is this procedure that underlies all applications).
• process monitoring (bearings, on-board computers, techno-processes, surveillance systems),
• processing textual and numeric information of large volume (especially!),
• understanding of the text and speech, and with the further development of technology and understanding of thoughts.

Additional scopes of technology (due to data compression), not related to EC. Any areas where you want to increase the bandwidth of the communication channel:

• data transfer,
• image transfer,
• voice transmission
• IP telephony

Based even on the listed properties, it can be concluded that the proposed technology is highly versatile. However, the technology is elegant and easy to implement. But despite the large list of advantages, the new approach does not claim to completely supplant traditional ES. Moreover, entire classes of tasks are preserved, where their application will soon remain much more effective than ES NP (for example, scheduling, planning, optimization, design, etc.)

AWP interface features

Interface Workstation. So far this is only a blank. (Only the properties are described. The technology is omitted whenever possible.) As the DB and BRs are required to be adapted to specific software, I consider it necessary to emphasize the role of the AWP, highlighting in a separate paragraph the description of the basic principles of the expert with the DB and KB. Here are only the distinctive features of the principles of operation and the user interface for ES NP in comparison with traditional ES.

What is there to start

The user of the NP NP is provided with a hierarchically structured network (an almost multi-partition graph, with arcs arranged according to a certain algorithm), which is a lexical-semantic software model. Each level of the hierarchy reflects the importance of the network element of this level in the software.

(The significance of a level is not a constant property of this level, but there is a statistical function of software. Nevertheless, the variation in significance of levels follows the distribution law, which is close to normal depending on the subject area).

The general tendency in the formed hierarchy of levels is as follows: the higher the level, the more significant in the software model are the elements of this level.

The main methods of work of the expert:

Here are only two basic methods of work, from which the listed properties of the ES follow. Along with them, in the course of interaction with the AWP, the expert will use other AWP methods.

As a result of the automatic structuring of software, each element of the constructed network reflects to some extent concepts that have circulation in software. The role of an expert is to change the significance of automatically proposed elements, enter descriptions of missing elements, clarify descriptions, and learn by yourself.

In the process, you can tune in to the work of the "upper level", i.e. deal with only a small number (10-20) of parameters known to the expert.

"Specify" and "Touch up" are two main methods of action. In a sense, they are reciprocal.

Principle of modularity

When the software implementation of ES NP uses the principle of modularity. The main modules in the technology of NP NP, which are the main difference from traditional ES, are the following:

• Module, the formation of the database and BR.
• The interface module workstation, consisting of two main modules:
• Interface module with DB and BR.
• Interface module with a description of the software.

The implementation of the remaining modules, which are presented in the diagram as "Other modules", is essentially no different from traditional ES. All the traditional subsystems of the ES, the need for which will arise when building an ES NP, can be attributed to the “Other Modules”.

Short summary

As can be seen from the description given, in this technology of work of an expert there is no need for an intermediate link, which, in the traditional approach to building an ES, was a knowledge engineer.

For the expert himself, the work is transferred to a qualitatively new level, giving him the opportunity to independently understand how he thinks and thinks. In other words, the expert is given the means "to dig into his own head."

In addition, after adaptation to the software of the generated database and knowledge base, the ES NP can work both fully automatically and automatically, being, for example, the main part in automatic decision-making systems when monitoring technological and other processes. Thus, ES NP can act as an independent application software.

The universality of the software model generation method makes ES NP the most promising tool for studying, classifying and structuring various software.

The same method allows you to automate the formation of a system of concepts and their signs (parameters) that previously appeared from various heuristic algorithms or as a result of long-term studies of a large group of workers. Now all this can be done by one expert, and the results will be the most objective, i.e. independent of the direction of the researcher’s thoughts.