WO2020153872A1

WO2020153872A1 - Method and system for searching for methods for resolving technical contradictions

Info

Publication number: WO2020153872A1
Application number: PCT/RU2019/050133
Authority: WO
Inventors: Галина Витальевна ПАСТУХОВА
Original assignee: Открытое акционерное общество "МБКИ"
Priority date: 2019-01-24
Filing date: 2019-08-19
Publication date: 2020-07-30
Also published as: RU2707917C1; EA202092614A1

Abstract

The invention relates to computer systems, and more particularly to systems for synthesizing intelligent solutions in the form of finding the necessary solution or a path thereto in response to a user query from a field of knowledge specified by the user. A system for implementing a method for searching for methods for resolving technical contradictions is configured in the form of a distributed hierarchical morphological structure which makes it possible to find, in response to a user query, a resolution to any technical contradiction in the form of finding solutions thereto with the aid of a system based on a trainable neural network which employs the principles of morphological analysis and fuzzy logic.

Description

METHOD AND SYSTEM FOR SEARCHING METHODS FOR RESOLVING TECHNICAL CONTRADICTIONS

DESCRIPTION

TECHNICAL FIELD OF THE INVENTION

The invention relates to computer systems, namely to systems synthesizing intelligent solutions in the form of finding the desired solution or a path to it from a given field of knowledge.

LEVEL OF TECHNOLOGY

Below are information about known technical means of a similar purpose as the proposed device.

Known system of semantic analysis and selection of documents with the ability to acquire knowledge, "DOCUMENT SEMANTIC ANALYSIS / SELECTION WITH KNOWLEDGE CREATIVITY CAPABILITY", see document US 2001/0014852. The system consists of a computer, a machine-readable knowledge carrier, a reader, a means of communication with local networks and the Internet, and an inference machine in the form of a separate semantic processor. The latter includes a block for preparing texts for modeling, a block for obtaining analyzed CAO structures, a block for synthesizing keywords / phrases for abstracts in natural language, a block for synthesizing abstracts, as well as a knowledge base block for CAO structures, a source code database block, a generalization database block natural language texts, a database block of new concepts in natural language, a database block of keywords and phrases for abstracts. The user, using accepted terms, enters the system with requests from the computer keyboard. The system simulates user requests in the form of CAO structures with the highlighting of keywords and phrases, on the basis of which it selects suitable candidate source texts in local databases and the Internet. CAO structures are also highlighted in candidate texts. If they correspond to the CAO-structures of user requests, then the candidate texts are stored for viewing in the database of natural language texts. Candidate texts that do not have the requested CAO structures are deleted. Based on the most frequent keywords, phrases and grammar rules, abstracts of selected texts are drawn up. It is possible to synthesize new concepts of knowledge based on the synthesis of selected CAO structures, taking into account the dictionaries of synonyms for the components of CAO structures.

The disadvantages of this system are as follows: - at each user request, the system sequentially considers the entire volume of available texts in external and internal networks, which requires significant expenditures of computer time;

- the capabilities of the system are limited by the available sources;

- splitting the sentences of the source texts into CAO-structures leads to the rupture of very complex semantic structures that are present inside the sentences and sections of the text;

- the system can only be used by a trained user expert.

Known knowledge management system for resolving situations under the RF patent

Ne2480826, which is designed to support the resolution of problem situations related to the unsatisfactory quality of specific objects. The knowledge management system for resolving situations contains a knowledge creation block, a knowledge organization block, to the input of which the output of the knowledge creation block is connected, a knowledge localization block, to the input of which the output of the knowledge organization block is connected, a knowledge positioning block, to the input of which the output of the knowledge localization block is connected, and a knowledge reuse unit, with inputs of which the outputs of the knowledge localization and positioning units are connected. The system integration block, the inputs and outputs of which are connected, respectively, to the inputs and outputs of the knowledge creation, knowledge organization, knowledge localization, knowledge positioning, knowledge reuse blocks, and the system visualization block, the input of which is connected to the system integration block output. The technical result consists in improving the characteristics of the processed information, as well as in improving the quality of visualization of situations.

Known intelligent information system for selecting knowledge "Optimel" under the RF patent NQ2564641, which provides the user with access to pre-systematized knowledge selected by analyzing the found sources of information, as well as ensuring the effective selection of the most appropriate method for solving the optimization problem. The technical result is to reduce the time for solving optimization problems. The indicated technical result is achieved by modeling the knowledge base using a binary tree of the system of questions and answers. This knowledge representation model makes it possible to simplify not only the implementation of the entire system, its use, but also the further filling of the base knowledge with new knowledge. The specificity of selection tasks requires periodic filling of the knowledge base with new knowledge, otherwise this system will cease to be relevant.

Known computer system for solving inventive problems "Inventing machine" (VM Tsurikov. Project "Inventing machine". Current state and real prospects. - Scientific and practical conference "Theory and practice of teaching technical creativity", Miass, 23- May 27, 1988. - pp. 37 - 43), which consists of a computer, a machine-readable medium, a software tool (PS) and a user and is designed to solve technical problems. It uses different knowledge models, combined into a knowledge base, and also uses an inference machine. This system searches for documents in local databases (DB) and on the Internet using semantic analysis. The latest version of the system is known as the Invention Machine Goldfire. See information booklet “Invention Machine Goldfire. Data sheet ", 2009).

It is used in a computer system - a computer - a computer-readable medium - a software tool (PS) - a user for solving intellectual problems for one direction - solving technical problems. It is a hybrid expert system (ES) using knowledge models combined into a knowledge base and having an inference engine. It is possible to search for suitable documents by queries based on semantic analysis, both in local databases (DB) and on the Internet. The "inventing machine" possesses knowledge (and produces answers) that in its part surpasses the knowledge of the most qualified specialists. It contains powerful information bases - technical special effects from various fields of knowledge, a catalog of methods for solving problems, some calculation models are used, in particular, functional and cost analysis. Some of these databases are built using knowledge models, in particular, semantic networks, production models, and subject-action-object models (CAO-models). Visual and graphic images are also used - graphs, images, drawings to improve work efficiency, connected to some knowledge models.

This technical solution, as the closest to the one declared in terms of the technical essence and the achieved result, was adopted as its prototype.

This system possesses knowledge of technical special effects from various fields of knowledge, is equipped with methods for solving inventive problems, uses visual and graphic images to improve work efficiency. The system was implemented on standard machine-readable media protected from copying by access codes without special devices for reading machine-readable media and their protection. It is connected to computers through reading devices, and during operation, the PS kernel is read into the computer and the work is performed in the computer's RAM with the pumping of information volumes from a computer-readable medium.

The disadvantages of this technical solution are:

- the system searches for documents in local databases (DB) and on the Internet using semantic analysis, but the user still has to independently work through the material found in order to select the appropriate option for solving a specific problem;

- ineffective ways of presenting data in databases and knowledge base are used, which do not allow to refuse unnecessary manipulations with data;

The “inventive machine” uses the technology of the theory of inventive problem solving (TRIZ), in which new technical solutions are sought using the same set of standard techniques for resolving technical contradictions.

DISCLOSURE OF INVENTIONS

The objective of the claimed invention is to create a technology that allows, at the user's request, to find a solution to any technical contradiction in the form of finding their solutions using a system based on a trained neural network that uses the principles of morphological analysis and fuzzy logic and forms stable relationships of the characteristics of the subject of the request, which is the basis for training neural networks and the basis of material for the formation of a detailed response to a request.

The essence of the first independent object of the claimed technical solution is expressed in the following set of essential features sufficient to solve the technical problem indicated by the applicant and obtain the technical result provided by the invention.

According to the invention, a method for finding methods for resolving technical contradictions, including analysis of known information about technical contradictions from different fields of technology, methods for solving inventive problems using logical analysis and issuing recommendations for a user's request, which is characterized by the fact that the user's request is entered into the system in language interface, recognize the interface language with the user's request and translate it into the request code, then create an algorithm for generating the request code, form priorities for training the neural network, then carry out additional training of the neural network according to the parameters corresponding to the request code, and search for methods for resolving technical contradictions with the help of a retrained neural network using the principles of morphological analysis and fuzzy logic, for which a group of basic constructive and socio-economic features is distinguished in the corresponding product or object request code, while the neural network forms stable relationships between the characteristics of the object corresponding to the request code with its social economic factors, then the neural network establishes the relationship between the subject under study and its mention in the patent databases, and each set of relationships is formed in its information space, after which, using the method of morphological design formations form a set of matrices, each of which is formed by orderly enumeration over an exhaustive and finite list of physical properties and socio-economic characteristics taken into account when applying technical solutions with the establishment of the type of dependence between them, after which, on the basis of the fixed relationships, a ranking system is created, which is the base to form a detailed response to a request, after which a response code is generated, translated into the interface language, and a response is issued to the system user.

In addition, the first independent object of the claimed technical solution is characterized by the presence of a number of additional optional features, namely:

- the request code is generated taking into account the input data filtering performed by the user;

- the response code to the user's request and the directions of the proposed solutions are specified by various types of filters.

The essence of the second independent object of the claimed technical solution is expressed in the following set of essential features sufficient to solve the technical problem indicated by the applicant and obtain the technical result provided by the invention.

According to the invention, a system based on a trained neural network for implementing the above-described method is made in the form of a distributed hierarchical morphological structure, including a request receiving module system user, a module for recognizing the interface language with a user request, a module for creating an algorithm for generating a request code, a module for forming a database structure corresponding to a request code, a module for generating priorities for training a neural network, a module for retraining a neural network using parameters corresponding to a request code, a neural network, searching for methods of resolving technical contradictions, a module for generating a response code, a module for translating the response code into the interface language and a module for issuing a response to a user.

The claimed set of essential features ensures the achievement of the technical result, which consists in the fact that the claimed method allows, at the user's request, to find a solution to any technical contradiction in the form of finding their solutions using a system based on a trained neural network using the principles of morphological analysis and fuzzy logic.

BRIEF DESCRIPTION OF THE DRAWINGS

The essence of the proposed technical solution is illustrated by the drawing in FIG. 1, which shows a block diagram of the claimed system.

DESCRIPTION OF THE PREFERRED EMBODIMENTS OF THE INVENTION

In the block diagram of FIG. 1 positions indicate:

1 - module for receiving a request from the system user, 2 - module for recognizing the interface language with a user request 3 - module for creating an algorithm for generating a request code, 4 - module for generating a request code, 5 - module for generating a database structure corresponding to the request code, 6 - module for generating priorities for training a neural network, 7 - a module for additional training of a neural network according to the parameters 1-p corresponding to the request code, 8 - a neural network, 9 - a module for generating a response code, 10 - a module for translating a response code into the interface language, 11 - a module for issuing a response to a user.

The system we declared works as follows.

Using module 1, a user request is entered into the system in the form of an interface text, the language of which is recognized by module 2. Then, module 3 creates an algorithm for generating a request code, and module 4 generates a request code. Module 5 generates a database structure corresponding to the request code, and module 6 generates priorities for training the neural network. Additional training of the neural network 8 according to the parameters corresponding to the request code is carried out using module 7. Next, the neural network 8 searches for methods for resolving technical contradictions with using the principles of morphological analysis and fuzzy logic. The morphological analysis method is based on combinatorics. Its essence lies in the idea of obtaining a detailed description of all existing and possible (admissible) technical systems of the class under study, followed by a search on this set for a description of the technical system that most fully corresponds to the set goal. For this, a group of basic structural or other features is distinguished in the product or object of interest. For each feature, alternative options are selected, that is, possible options for its implementation. By combining them with each other, many different solutions are obtained, including those of practical interest. Preliminarily, neural network 8 forms stable interconnections of the characteristics of the subject corresponding to the request code with its socio-economic factors in the form of layering of pseudo-planes, learns on the obtained interconnections and searches for a solution to overcome this or that technical contradiction or technical problem and specifies the direction of their solution with various types of filters, then the neural network establishes the relationship between the object under study and its mention in the patent databases and each set is formed in its information space, after which, using the method of morphological design, a set of matrices is formed, each of which is formed by orderly enumerations over an exhaustive and finite list of physical properties and socially - economic characteristics taken into account when applying technical solutions with the establishment of the type of dependence between them, after which, on the basis of the fixed relationships, work is carried out with even coy and fuzzy logic, create a ranking system, which is the basis for training a neural network and the basis of material for the formation of a detailed response to a request. Further, module 9 generates a response code, module 10 translates the response code into the interface language, and module 11 provides a response to the user.

The claimed device can be implemented using known equipment, technical and technological means in the form of well-known universal neural network packages for modeling a wide range of artificial neural networks, managers of trained artificial neural networks, etc.

Claims

CLAIM

1. A method of searching for methods for resolving technical contradictions, including the analysis of known information about technical contradictions from different fields of technology, methods for solving inventive problems using logical analysis and issuing recommendations to the user's request, characterized in that:

enter the user's request into the system in the interface language,

recognize the interface language with the user's request and translate it into the request code, then

create an algorithm for generating the request code,

form priorities for training the neural network, after which

carry out additional training of the neural network according to the parameters corresponding to the request code, and

search for methods for resolving technical contradictions using a retrained neural network using the principles of morphological analysis and fuzzy logic, for which a group of basic constructive and socio-economic features is identified in the corresponding product or object request code, while the neural network forms stable interconnections of the characteristics of the corresponding the request code of the subject with its socio-economic factors, then the neural network establishes relationships between the subject under study and its mention in the patent databases, and each set of relationships is formed in its information space, after which

using the method of morphological design, a set of matrices is formed, each of which is formed by orderly enumeration of the exhaustive and final list of physical properties and socio-economic characteristics taken into account when applying technical solutions with the establishment of the type of relationship between them, after which

based on the fixed relationships, a ranking system is created, which is the basis for the formation of a detailed response to a request, after which

generate a response code, translate it into the interface language, and

give a response to the system user.

2. The method according to claim 1, characterized in that the request code is generated taking into account the filtering of the input data made by the user.

3. The method according to claim 1, characterized in that the response code to the user's request and the directions of the proposed solutions are specified by various types of filters.

4. A system based on a trained neural network for implementing the method according to the previous paragraphs, made in the form of a distributed hierarchical morphological structure, including:

a module for receiving a request from a system user,

interface language recognition module with user request,

a module for creating an algorithm for generating a request code,

a module for generating a database structure corresponding to the request code, a module for generating priorities for training a neural network,

a module for additional training of a neural network using the parameters corresponding to the request code, a neural network that searches for methods for resolving technical contradictions,

response code generation module,

a module for translating the response code into the interface language, and

module for issuing a response to the user.