RU2692043C2 - Method for automatic classification of formalized text documents and authorized users of electronic document management system - Google Patents

Abstract

FIELD: computer equipment.SUBSTANCE: method includes: extraction of metadata and informative part of document, conversion of document from storage format into text, conversion of words into word forms, discarding non-significant words, counting word weights, generating a set of classification features, wherein at the training step, a system of predicates for identifying the confidentiality mark of the document is generated based on the set of classified documents; at the document classification step, based on the characteristics, a decision is made on the relevance of the document of each of the confidentiality marks, at the training stage, based on the set of manually classified authorized users, forming a predicate identification system of their confidentiality mark, wherein on the basis of confidentiality marks of incoming documents and access rights of authorized users of system to these documents form a set of classification features.EFFECT: automatic classification of formalized text documents and authorized users of electronic document management system according to confidentiality marks.1 cl, 1 dwg, 1 tbl

Description

The invention relates to document classification systems and can be used in electronic document management systems, databases, automated systems using confidentiality tags, where the access objects are formalized text documents, the subjects are authorized users, under conditions of an arbitrary number of confidentiality marks used.

The level of technology

a) Description of analogues

A well-known analog is the method of automatic classification of documents (Li Y., Jain A. "Classification of text documents", The Computer Journal 41, 8, pp. 537-546, 1998), which consists in converting a document from the storage format to text in natural language, translate words of a document into basic word forms, discard insignificant words, calculate weights of words in a document in accordance with the frequencies of their occurrence, at the training stage, according to the presented set of manually classified documents, form a set of classification features, and document fictions convert it from storage format to natural language text, transform document words to basic word forms, discard insignificant words, calculate word weights in a document, determine whether a document belongs to an information area based on simple Bayesian classification criteria [1] .

The disadvantages of this method are:

does not allow to classify formalized text documents by confidentiality marks;

does not allow to classify authorized users by confidentiality tags.

An analogue is also known - the method of automatic classification of documents (Pat. 6327581 United States of America, IPC G06F 015/18. [Text] / CarltonJ .; applicant and patent holder Microsoft Corporation. - №09 / 055477 ; declared April 6, 1998; publ. December 4, 01), which consists in converting a document from the storage format into natural language text, converting the words of the document into basic word forms, discarding insignificant words, counting the word weights in the document in accordance with with the frequencies of their appearance; at the training stage, a set of classification signs is formed according to the presented set of manually classified documents; when a document is classified, it is converted from storage format into text in natural language, the words of the document are converted into basic word forms, the insignificant words are rejected, the word weights are calculated in the document The SVM (SupportVectorMachines) criterion and classification criteria determine the document's belonging to the information area [2].

The disadvantages of this method are:

does not allow to classify formalized text documents by confidentiality marks;

does not allow to classify authorized users by confidentiality tags.

An analogue is also known - a multiclass classification method (Schapire RE, Singer Y. "BoosTexter: A boosting-based system for text categorization". MachineLearning 39, 2/3, 2000, pp. 135-168), which consists in converting document from the storage format in the text in natural language, convert the words of the document into basic word forms, discard insignificant words, calculate the weights of words in the document in accordance with the frequencies of their appearance and thereby form a feature vector of the document, at the stage of training according to the presented set of manually classified documents ntov form a set of classification features, retain classification features in the database, when classifying a document, convert it from the storage format to natural language text, convert the words of the document into basic word forms, discard insignificant words, calculate the weights of words in the document and form a vector of document features , then decide on the ownership or non-ownership of the document of each of the information area [3].

The disadvantages of this method are:

does not allow to classify formalized text documents by confidentiality marks;

does not allow to classify authorized users by confidentiality tags.

An analogue is also known - a method of automatic classification of documents (Pat. 2254610 Russian Federation, IPC G06F 17/30. Method of automatic classification of documents [Text] / Agranovsky AV, Arutyunyan R.E., Khadi R.A., Telesnin B. A.; Applicant and patentee State Scientific Institution Scientific Research Institute "SPETSVUZAVTOMATIKA". - №2003126907/09; declared September 4, 03, published on June 20, 2005), which is transforming the document from the storage format into text on natural language transform the words of the converted document the basic word forms, discard insignificant words, calculate the weights of words in the above-mentioned document in accordance with the frequencies of their appearance and thereby form a feature vector of the document; at the training stage, a set of classification features is stored in the database, when classifying a document, it is transformed from a storage format into text in a natural language, transform the words of a document into basic word forms , discard insignificant words, calculate the weights of words in the document and form a vector of signs of the document, after which they decide whether or not the document belongs to each category [4]. The disadvantages of this method are:

does not allow to classify formalized text documents by confidentiality marks;

does not allow to classify authorized users by confidentiality tags.

An analogue is also known - a method of automatic classification of formalized documents in an electronic document management system (Pat.2546555 Russian Federation, IPC G06F 17/30. Method of automatic classification of formalized documents in an electronic document management system [Text] / SV Nosenko, ID Korolev, Poddubny MI; applicant and patent holder Federal State State Military Educational Institution of Higher Professional Education "Marshal of the Soviet Union Communications Academy named after SM Budyonny" of the Ministry of Defense of the Russian Federation - No. 2013155168/08; declared December 11, 2013; published on April 10, 2015), which defines the areas of a formalized document for extracting metadata and the informative part, converts the document from the storage format to text on natural language, transform the words of the converted document into basic word forms, discard insignificant words, calculate the weights of words in the document in accordance with the frequencies of their appearance and thereby form the features of the document: at the learning stage poo of manually classified documents form a set of classification signs, save classification signs in the database; when classifying a document on the basis of the received classification features using a database, decide on the relevance of the document to each of the information areas, at the stage of determining the ownership of the document from each information area using a priori information about the category dependencies from each other [5].

The disadvantages of this method are:

does not allow to classify formalized text documents by confidentiality marks;

does not allow to classify authorized users by confidentiality tags.

b) Description of the closest analogue (prototype)

The closest in technical essence to the proposed is a method for automatic classification of confidential

formalized documents in the electronic document management system, which consists in defining the areas of the formalized document for extracting metadata and informative part, converting the document from the storage format into text in natural language, converting the words of the converted document into basic word forms, discarding insignificant words, calculating weights words in the document in accordance with the frequency of their appearance and thereby form the characteristics of the document, at the stage of training on a set of classifications manually generated documents form a set of classification signs, save classification signs in the database, when classifying a document based on the received classification signs using a database, decide whether the document is related to each of the information areas and to each of the privacy marks specified in the information system the stage of determining the document’s belonging to each information area and privacy label is used a priori information about dependency yah categories apart [6].

The disadvantages of this method are:

does not allow to classify formalized text documents by confidentiality labels in relation to systems in which the number of confidentiality labels is arbitrary, which significantly complicates implementation;

does not allow to classify authorized users by confidentiality tags.

Disclosure of the invention

The aim of the present invention is to provide automatic classification of formalized text documents and authorized users of the electronic document management system by confidentiality tags.

The technical result is achieved by calculating the areas of a formalized document with the computing device to extract the metadata and informative part, converting the document from the storage format into natural language text, converting the words of the converted document into basic word forms, discarding insignificant words, counting the weights of the words in the document according to the frequencies of their occurrence, they form a set of classification signs: at the stage of training on a set of classified hands th documents form a system of identification marks of predicates confidential documents, retain the system of predicates in the database; at the stage of document classification, on the basis of received classification criteria, a decision is made on the relevance of the document to each of the confidentiality marks, characterized in that based on the confidentiality marks of the received documents and the access rights of authorized users of the system to these documents, extracted from the matrix of access rights, set of classification features: at the stage of training on a set of manually classified authorized users form a system of predicates and identifying their privacy labels and storing the predicate system in a database; at the stage of classification of an authorized user, the resulting set of classification signs is substituted into the predicate system located in the database, according to the predicate that has assumed the value “true”, decide on the relevance of the authorized user to one of the confidentiality marks.

- множество ключевых слов

реквизита, где

q' - количество всех используемых ключевых слов.This technical result is achieved due to the fact that carry out the selection of the characteristics of the same parts of the text Z - details. At the same time, the number of different details of a formalized document is limited [7]; in addition, some of them do not determine the individuality of the document form (characteristic of all forms or not used in these conditions). Each attribute is expressed by the final predicate P _Z (T, L), where T is a finite set of text characteristics t,

- many keywords

props where

q 'is the number of all keywords used.

The rule for constructing a predicate for recognizing the prop of a formalized document is expressed by the following formula [8]:

- предикат узнавания значения а h-той переменной текста;Where

- predicate of recognition of the value of a h-that text variable;

μ is the number of text variables;

ϕ - the value of the alphabet of the h-text variable;

- предикат узнавания значения b ключевого слова q соответствующего i-той зоне.

- predicate of recognition of the value b of the keyword q corresponding to the i-th zone.

- множество форм документа; j' количество всех используемых форм документов, Z - множество реквизитов документа, n - количество всех реквизитов документов,

- множество ключевых слов, где

q' - количество всех используемых ключевых слов.The form of the document is expressed by the final predicate P _V (Z, L), where V = {ν _j }, where

- many forms of the document; j 'is the number of all document forms used, Z is the set of document details, n is the number of all document details,

- many keywords, where

q 'is the number of all keywords used.

The rule for constructing a predicate for recognizing the form of a document is expressed by the following formula [8]:

z_i - предикат узнавания реквизита для j-той формы документа;Where

z _i - predicate recognition predicate for the j-th form of the document;

- предикат узнавания уникального значения γ ключевого слова q j-той формы документа.

- the predicate of recognition of the unique value γ of the keyword q of the j form of the document.

Using the rules (1, 2), predicate identification systems and document forms are created.

The form of the document uniquely specifies the location of the details of the document, which allows to classify documents according to the form of the document and determined by the corresponding access restriction.

Then the informative part of the document (hereinafter referred to as the text) is transformed from the storage format into text in natural language, the words of the document are converted into basic word forms, the insignificant words are discarded, the weights of words in the text are calculated in accordance with the frequencies of their appearance, and thus the predicates of identifying text characters are formed .

In the training mode, according to the presented set of hand-classified texts, a predicate identification system of text features is formed, where the number of predicates in the predicate system is determined by the number of areas of information responsibility for which the documents must be classified (the number of executors in the automated system). Store predicates in the database.

β' - количество областей информационнойThe rule for constructing the predicate P _U (W) of recognition of the information area U = {u _β }, where

β '- the number of areas of information

responsibility, is expressed by the following formula [8]:

where W = {w _p } is the set of significant words of the texts, where

p '- the number of significant words of texts;

- предикат узнавания значения веса ƒ значимого слова w_p, в тексте документа

информационной области по g-тому значению веса слова.

- predicate of recognition of the value of weight ƒ significant word w _p , in the text of the document

information area for the g-th word weight value.

The rule of formation of the system of predicates of recognition of the document confidentiality label implies the number of information areas comparable to the number of applied confidentiality labels in the system [6]. To implement the method in more complex information systems it is proposed to apply the algebra of finite predicates.

узнавания метки конфиденциальности М={m_λ}, где

λ' - количество определенных в системе меток конфиденциальности выразится следующей формулой [10]:Using the proposed algebra [9], a rule for constructing a system of predicates is presented that does not require additional transformations for any information system. Thus, the predicate construction rule

recognition of the privacy label M = {m _λ }, where

λ '- the number of confidentiality labels defined in the system will be expressed by the following formula [10]:

- предикат узнавания k-то значения r-ого реквизита;Where

- the predicate of recognition of the k-value of the r-th props;

m ₀ is the privacy label of the object (document d _y ), with m _o ∈ M;

- предикат узнавания β-той области, где

β' - количество информационных областей системы.

- the predicate of recognition of β-the area where

β 'is the number of information areas of the system.

After determining the confidentiality marks of the information system objects, the classifier proceeds to the classification stage of a finite set of subjects (authorized users) S. Based on the matrix of access rights and confidentiality marks, a set of classification indicators is formed, including the confidentiality marks of all objects to which the subject has access rights (let's call these objects are analyzed), the current label of confidentiality of the classified entity, if available, and the data provided tions of the security environment of the functioning of the classifier [11, 12].

The classifier operation is explained on the example of an electronic document management system, in which four confidentiality marks are specified.

x_s с величиной алфавитов 4, 4, 2 соответственно, гдеIn order to develop a rule for constructing a system of predicates of recognition of the label of privacy of subjects, we introduce a set of variables m _o

x _s with the value of alphabets 4, 4, 2, respectively, where

m _o - privacy label of the object being analyzed;

- текущая метка конфиденциальности классифицируемого субъекта;

- current confidentiality label of the subject being classified;

x _s - security conditions for the classification of the subject of the electronic document management system.

The structure of the features used in this example is conveniently presented in the form of table 1.

The unambiguity and correctness of the classification of the subject for λ = 4 confidentiality marks are proved by brute force. The proof is similar for any λ. At the same time, for the reaction of the system to the label assignment, it is not the presence of the necessary conditions that is checked, but the absence of interfering ones, since the subject does not have the current label of the subject being classified in the flow analysis [13].

Applying the algebra of finite predicates, we construct a system of predicates for determining the confidentiality label of a classified subject [9]:

The presence of an “additional” predicate (preservation of the current state), which is not peculiar to the rules (1-4), is a necessity. It eliminates the constant triggering of the training mode when a security policy violation is detected, automatically making an a priori given decision. In the claimed method, this ignoring a potentially dangerous object to be analyzed. This approach eliminates the overestimation of the subject confidentiality label.

выражается формулой:Thus, the rule for constructing a predicate for recognizing the confidentiality label of subjects

expressed by the formula:

- предикат узнавания значения метки конфиденциальности Я анализируемого объекта d_y;Where

- predicate of recognizing the value of confidentiality label I of the analyzed object d _y ;

- предикат узнавания текущей метки конфиденциальности λ классифицируемого субъекта s;

- the predicate of recognition of the current label confidentiality λ of the classified entity s;

- предикат узнавания соблюдения условий τ безопасной классификации субъекта s.

- predicate of recognition of compliance with the conditions τ of a secure classification of a subject s.

Classification security conditions are formulated within the framework of an organization’s security policy and may differ significantly depending on the purpose of the classifier and the value of the information processed in the system. The implementation of a security policy in a classified information system requires a separate study and, in the framework of the proposed method, is limited to a safe-unsafe scheme.

In the training mode, based on the presented set of manually classified objects and subjects of the electronic document management system, predicate systems of identification of their confidentiality marks are formed. The number of predicates in the system is determined: for objects, the number of confidentiality labels specified in the information system, for subjects - the number of confidentiality labels +1. Store predicate systems in a database.

In the system operation mode, at the stage of classification of objects (documents), the document is converted from a storage format into natural language text, text words are converted into basic word forms, insignificant words are discarded, word weights are counted in the text, the resulting values are substituted into the predicate system (4) in the database. According to the predicate, which accepted the truth value “1”, the area of information responsibility and the confidentiality label of the classified object (document) is determined.

When classifying subjects of an information system, a set of classification features is formed for each of them. These characteristics are extracted from the matrix of access rights and include: a set of the set of analyzed objects, their confidentiality marks, current confidentiality marks of the classified entity (if any) and data on security policy The set of features determined in this way is substituted into the predicate system (5) in the database. According to the predicate that accepted the truth value “1”, the confidentiality label of the classified entity is determined.

In this case, if it is necessary to use a priori information about the dependencies of the confidentiality labels from each other, we use the algebra of finite predicates [9], which allows us to carry out a full range of operations on logical expressions. Note that this method is intended for processing machine-readable texts in natural language.

Weight ƒ w _p word forms in the text of the document d _y , is calculated by the formula:

- количество раз, которое w_p-я словоформа встречается в d_y-м тексте документа,

- общее количество словоформ в d_y-m тексте документа.Here

- the number of times that the w _p th word form occurs in the d _y th text of the document,

- the total number of word forms in the d _y -m text of the document.

Documents for classification can be presented in various formats, allowing the selection of text content from them. These can be text files of various formats, graphic files with a graphic representation of some text, sound files with speech recording and other files for which there is a mechanism for extracting text from them that reflects their content. In addition, the classifier should be able to extract information from the matrix of access rights, regardless of its implementation (object properties, subject profile or system matrix).

Each object (either training or subject to classification) passes through a preprocessing stage, which determines the format of the document and determines whether it is possible to extract text from a document of this format. In the case of a positive decision, the text is extracted from the document. After splitting the text into words, the definition for each word of its base word form occurs in one of the ways [14-17]. Most often, the Porter algorithm is used to solve such problems, which consists in using special rules for cutting off and replacing word endings.

According to the proposed method, each object d _y is represented by a Cartesian product of variables from the T × L × W sets, where the classifier training stage serves to initialize the classifier and build the classification criteria. This should be given a set of training documents, manually classified in advance by the labels of confidentiality. After extracting textual content from them, a dictionary of meaningful words is constructed. The dictionary contains the basic word forms of all words found in the training documents.

When classifying a document, not all word forms from the dictionary of documents are taken into account, but only those that are included in the working dictionary of the classifier of this privacy label. The most informative word forms from the point of view of determining the ownership of a document in this category, not included in the stop dictionary, are included in the classifier's working dictionary. The informativeness of the word form w _p for the classifier for the information area u _β is determined by the following formula [18]:

At the same time, the threshold of informativeness ε is set all word forms that do not fall into the stop dictionary, the information content of which exceeds this threshold, are included in the classifier’s work dictionary. The stop dictionary consists of word forms whose frequency of occurrence in a set of training documents exceeds a predetermined threshold δ. This cut off words that do not carry a semantic load, such as prepositions, conjunctions, introductory and common words, etc. The values of the coefficient δ, according to this method, are set in the range from 0.05 to 0.7, depending on the specific use of the method. The values of the threshold of informativeness δ can be different in different conditions of use of the method.

By unambiguously identifying the area of information responsibility to which the document of (3) belongs, using the extracted document metadata, namely, the document details defining its confidentiality obtained by (1), the corresponding confidentiality mark is determined. For this, the indicated values are substituted into the system of predicates constructed by (4). According to the predicate that accepted the truth value "1", the confidentiality label is determined from the list.

Having determined the confidentiality marks of the information system objects, the classifier proceeds to the stage of classification of subjects. For this, sets of classification signs for each analyzed object, extracted from the matrix of access rights, are substituted into the predicate system. The predicate system is built at the stage of learning by the rule (5) and is in the database. After analyzing all the objects with respect to the subject being classified, the prediction mark is determined by the predicates that accepted the truth value “1”.

Comparative analysis of the proposed solution with the prototype shows that the proposed method differs from the well-known formation at the training stage and the application at the stage of classification of the predicate system, built according to rule (5). As well as an improved rule (4).

Thanks to a new set of essential features, the method enables the classification of subjects of the electronic document management system according to confidentiality marks.

The analysis of the level of technology has allowed to establish that the analogues, characterized by a set of features that are identical to all the features of the claimed technical solution, are absent, which indicates the compliance of the claimed method to the condition of patentability "novelty."

Search results known solutions in this and related areas of technology in order to identify signs that match the distinctive features of the prototype features of the declared object, showed that they do not follow explicitly from the prior art. The prior art also revealed no prominence of distinctive essential features causing the same technical result, which was achieved in the claimed method. Therefore, the claimed invention meets the condition of patentability "inventive step".

Brief Description of the Drawings

The figure shows a block diagram of a computing device for implementing the method.

The device for implementing the method consists of the following blocks: source of documents 1, analyzer of text characteristics 2, recognition of the details of document 3, recognition of the form of document 4, selection of metadata 5, definition of basic word forms 6, creation of a working dictionary 7, determination of weights of word forms of the text of the document 8, area recognition information responsibility 9, accounting of document on metadata 10, training 11, recognition of the confidentiality label of the object (document) 12, addressing of document 13, recognition of the confidentiality label from 14, loading into the system in accordance with the received classification 15. Implementation of the invention

Automatic classification of formalized text documents and authorized users of the electronic document management system is as follows:

1. In the classification mode.

в них. Значения t и

участков документа поступают в блок 3, где с помощью системы предикатов, построенных по правилу (1) распознаются реквизиты документа. Информация о распознанных реквизитах документа поступает в блок 4, где система предикатов, построенная по правилу (2) осуществляет распознавание формы документа.At the stage of classification of objects (formalized text documents). When a new document appears in the document source 1, it enters block 2, which reveals the values of the text characteristics of the t document sections and keywords

in them. Values of t and

plots of the document are received in block 3, where using the system of predicates built according to rule (1), document details are recognized. Information about the recognized details of the document goes to block 4, where the predicate system built according to rule (2) recognizes the form of the document.

In block 5 of the incoming document from block 2, using information about a specific form of the document from block 4, which uniquely specifies the location of the document details, the required values of the details, which are the metadata of the document. The document and the corresponding metadata come in block 10, where the document is taken into account according to its metadata and the storage of its reference copy is organized. Also uniquely defined in block 5, the informative part of the document goes to block 6, where the words are converted into word forms. Received in block 6 word forms come in block 7, where in the process of the system is the creation of a working dictionary of significant words.

путем вычисления значений предикатов системы предикатов, построенной по правилу (3).The word forms obtained in block 6 are transferred to block 8, where weights ƒ word forms of the informative part of the document that fall into the working dictionary are calculated. From block 8, the weights of the received word forms go to block 9, where the information area is recognized

by calculating the predicate values of the predicate system constructed according to rule (3).

From block 10, the document goes to block 13, and the metadata goes to block 12 and 13. In block 12, based on the metadata received from block 10, namely the document details, and the values obtained in block 9, based on the predicate system built according to the rule (4 ) A confidentiality label corresponding to the classified document is determined. In block 13, the received document and metadata from block 10, based on the values received from blocks 12 and 9, are assigned the corresponding access rights and confidentiality label of the classified document.

Next, through block 15, the document is loaded into the information system in accordance with certain classes, (prescribing access rights and confidentiality marks).

At the stage of the classification of subjects (authorized users). Block 14 of Block 15 receives data on the objects being analyzed (confidentiality labels of documents to which the classified entity has read, write access rights; data characterizing compliance with the security policy) and the current confidentiality label of the classified entity. In block 14, based on the data received from block 15 and the predicate system built according to rule (5), a confidentiality label corresponding to the subject being classified is determined.

It is possible to adjust the labels

classified subjects when new documents enter the system. Then the privacy label of the object being analyzed comes to block 14 from block 13.

Further, through block 15, the confidentiality label obtained in block 14 is loaded into the information system. 2. In training mode.

The system learning mode is used in the following cases:

В этом случае оператором системы через блок 11 вносятся изменения в систему предикатов блока 3 или определяется реквизит документа «вручную»;if it is impossible for the system to recognize the predicates of the document details in block 3 by the values of the document variables t and

In this case, the system operator, through block 11, makes changes to the predicate system of block 3 or determines the document attribute “manually”;

if it is impossible for the system to recognize the predicates of the document form in block 4, based on the predicates of the predicate system of block 3. In this case, the system operator through block 11 makes changes to the predicate system of block 4 or determines the document form "manually";

if it is impossible for the system to recognize the predicates of the information area in block 9 by the values of the weights of significant words from the working dictionary extracted from the informative part of the document. In this case, the system operator through block 11 makes changes to the predicate system of block 9 or determines the information area of the document "manually";

if it is impossible for the system to recognize the predicates of the object's confidentiality mark in block 12, the predicate values of the predicate system of block 9 and the block 10 metadata. In this case, the system operator modifies the predicate system of block 12 through block 11 or determines the confidentiality mark "manually";

if it is impossible for the system to recognize the predicates of the subject's confidentiality mark in block 14 according to block 15. In this case, the system operator makes changes to the predicate system of block 14 through block 11 or determines the confidentiality mark "manually".

Thus, the method allows to classify not only objects of the electronic document management system (formalized electronic text documents) by confidentiality marks, but also the subjects of the system (authorized users, performers), thus achieving the above technical result.

Information sources:

1. Li Y., Jain A. "Classification of text documents", The Computer Journal 41, 8, pp. 537-546, 1998.

2. Pat. 6327581 United States of America, IPC G 06 F 015 / 18. Metals and apparatus for the support vector machine [Text] / CarltonJ .; applicant and patent holder of Microsoft Corporation. - №09 / 055477; declare 04/06/98; publ. 12.12.01.

3. Schapire R.E., Singer Y. "BoosTexter: A boosting-based system for text categorization". MachineLearning 39, 2/3, 2000, pp. 135-168.

4. Pat.2254610 Russian Federation, IPC G 06 F 17/30. Method of automatic classification of documents [Text] / Agranovsky AV, Arutyunyan R.E., Hadi R.A., Telesnin B.A .; applicant and patent holder State Scientific Institution Research Institute "SPETSVUZAVTOMATA". - № 2003126907/09; declare 09/04/03; publ. 06/20/05.

5. Pat. 2546555 Russian Federation, IPC G06F 17/30. The method of automatic classification of formalized documents in the electronic document management system [Text] / Nosenko SV., Korolev I.D., Poddubny M.I .; Applicant and patent holder Federal State State Military Educational Establishment of Higher Professional Education “Marshal of the Soviet Union Communications Academy named after Marshal of the Soviet Union S.M. Budyonny ”of the Ministry of Defense of the Russian Federation. - # 2013155168/08; declare 11/12/2013; publ. 04/10/2015.

6. Application for the invention №2015152418 from 07.12.2015. The method of automatic classification of confidential formalized documents in the electronic document management system [Text] / M.I. Poddubny, I.D. Korolev, SV; Applicant and patent holder Federal State Official Military Educational Institution of Higher Professional Education “Krasnodar Higher Military School named after Army General S.М. Shtemenko ”of the Ministry of Defense of the Russian Federation - №2015152418; declare 12/07/2015; publ. 13.06.2017 Bull. №17.

7. GOST R 6.30-2003. Unified documentation systems. Unified system of organizational and administrative documentation. Requirements for the paperwork: approved. and enter. into effect by the Resolution of Gosstandart of Russia dated 03.03.2003 No. 65 - art. - M .: Standardinform, 2007 - 17 p.

8. Korolev I.D. Approaches to the operational identification of formalized electronic documents in automated records management / I.D. Korolev, SV Nosenko // Polythematic network electronic scientific journal of the Kuban State Agrarian University (Scientific journal of KubSAU) [Electronic resource]. - Krasnodar: KubSAU, 2013. - №08 (092). - IDA [article ID]: 0921308074. -Access mode: http://ej.kubagro.ru/2013/08/pdf/74.pdf, 0.875 u.s.

9. M.F. Bondarenko, Yu.P. Shabanov-Kushnarenko. On the algebra of finite predicates. [Text] // Scientific and technical journal "Bionics of intelligence". KNURE, Kharkov, Ukraine - 2011 No. 3 (77).

10. Method for automatic classification of confidential formalized documents in the electronic document management system / D.V. Malyshev, I.N. Shaykov, M.I. Poddubny, I.D. Korolev // Telecommunications .: every month. Inform., analyte. and study.-method, journal, Moscow: MSTU. Bauman, 2016 №8. - p. 18-22.

11. “Guidance document. Automated systems. Protection against unauthorized access to information. Classification of automated systems and information security requirements ”(approved by the decision of the State Technical Commission of Russia of March 30, 1992).

12. “Guidance document. Computing facilities. Protection against unauthorized access to information. Indicators of protection against unauthorized access to information "(approved by the decision of the State Technical Commission of Russia on March 30, 1992).

13. Methods of automatic classification of entities of the electronic document management system according to confidentiality marks // Internet magazine “Technosphere Safety Technologies” (Scientific Journal of the Academy of the State Fire Service) [Electronic resource]. - M .: 2016. - №6 (70). - ISSN 2071-7342. - access mode http: //agps-2006/narod.ru//ttb/2016-6/29-06-16.ttb.pdf.

14. Porter M.F. "An algorithm for suffix stripping", Program, Vol. 14, No. 3, 1980, pp. 130-137.

15. Pat.2096825 Russian Federation, IPC G06F 17/00, G06F 17/30. Information processing device for information retrieval [Text] / Kovalev, MV, Virgunov, IV, Naimushin, IA, Chetverev, VV; applicant and patent holder Limited Liability Company "Information Bureau". - №96119820 / 09; declare 10.14.96; publ. 11/20/97, Bull. №14.

16. Pat. 6308149 United States, IPC G06F 17/27. Grouping words with equivalent substrings / Gaussier E., Grefenstette G., Chanod J.-P .; applicant and patent holder Xerox Corporation. - №09 / 213309; declare 12/16/98; publ. 23.10.01.

17. Pat. 6430557 United States of America, IPC G06F 017/30; G06F 017/27; G06F 017/21. Identifying a group of words using a modified suffix relationships / Gaussier E., Grefenstette G., Chanod J.-P .; applicant and patent holder Xerox Corporation. - №09 / 212662; declare 12/16/98; publ. 08.08.02

18. Craven M., DiPasquo D., Freitag D. et al. "Learning to construct knowledge bases from the World Wide Web", Artificial Intelligence, Vol. 118 (1-2), 2000, pp. 69-113.

Claims (1)

The method of automatic classification of formalized text documents and authorized users of the electronic document management system, consisting in that the computing device determines the areas of the formalized document for extracting metadata and informative part, converts the document from the storage format into text in natural language, converts the words of the converted document into basic word forms , discard insignificant words, calculate the weights of words in the document in accordance with Vii with the frequencies of their appearance and form a set of classification features: at the stage of training on a set of manually classified documents form a system of predicates for identifying the document confidentiality label, save the system of predicates in the database; at the stage of document classification, on the basis of received classification criteria, a decision is made on the relevance of the document to each of the confidentiality marks, characterized in that based on the confidentiality marks of the received documents and the access rights of authorized users of the system to these documents, extracted from the matrix of access rights, set of classification features: at the stage of training on a set of manually classified authorized users form a system of predicates and identifying their privacy labels and storing the predicate system in a database; at the stage of classification of an authorized user, the resulting set of classification signs is substituted into the predicate system located in the database, according to the predicate that has assumed the value “true”, decide on the relevance of the authorized user to one of the confidentiality marks.

Images