Eng | Rus

“Informatics and Applications” scientific journal

Volume 14, Issue 4, 2020

Content   Abstract and Keywords   About Authors

ON PROBABILISTIC ESTIMATES OF THE VALIDITY OF EMPIRICAL CONCLUSIONS
  • A. A. Grusho  Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • M. I. Zabezhailo  A. A. Dorodnicyn Computing Center, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 40 Vavilov Str., Moscow 119333, Russian Federation
  • D. V. Smirnov  Sberbank of Russia, 19 Vavilov Str., Moscow 117999, Russian Federation
  • E. E. Timonina  Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: The work focuses on some features of data analysis in insider search problems. The possibilities of using different approaches to describe the diagnosis of insider actions in the analysis of large empirical data are discussed.
In tasks of this type, it is necessary to establish (predict, diagnose, etc.) the presence or the absence of target properties in any users from a given set. The assessment of the correctness of plausible reasoning is checked on the basis of estimates of the probabilities of the random appearance of the found laws in the simplest probabilistic models. The examples discussed show at what ratios of parameters it is possible to effectively identify correlations between events with which insiders can be identified. Two methods of controlling relations between parameters are indicated, allowing to obtain content information. The first method is based on dividing the observation period at the intervals during which the desired correlation may appear. The second method relates to the ways to reduce the set of users that could potentially become insiders, i. e., the authors are talking about the formation of clusters in which probabilistic estimates become operational. The desired relationships between the parameters for finding correlations can be determined using limit theorems in the series scheme.

Keywords: hostile insider; causal analysis; probabilistic estimates of random appearance of properties

IMPACT OF THE ISOLATION PARAMETERS ON RESOURCE ALLOCATION IN THE NETWORK SLICING MODEL
  • F. A. Moskaleva  Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation
  • Yu. V. Gaidamaka  Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation, Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • V. S. Shorgin  Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: Network slicing technology is defined as one of the main components of the fifth generation of mobile communications, capable of solving the problem of the colossal growth of data traffic in cellular networks. A key feature of slicing that limits the impact of one slice on another is to provide isolated performance guarantees to deliver high quality of service. In this article, a model of resource allocation during slicing is developed using the queuing theory. The main task of the work is to determine how network resources should be fairly shared between two slices in the system. The resource allocation problem is formulated as an optimization problem. For the constructed model, a numerical analysis was carried out showing the significant effect of the isolation parameters on the performance characteristics of the system.

Keywords: network slicing; fairness resource allocation; isolation of slices; isolation parameter

RETRIAL QUEUING MODEL FOR ANALYZING JOINT URLLC AND eMBB TRANSMISSION IN 5G NETWORKS
  • P. A. Kharin  Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation
  • E. D. Makeeva  Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation
  • I. A. Kochetkova  Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation , Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • D. V. Efrosinin  Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation , Johannes Kepler Universitaet Linz, 69 Altenbergerstrasse, Linz 4040, Austria
  • S. Ya. Shorgin  Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: Fifth-generation (5G) networks are characterized by three use cases - massive machine-type communication (mMTC), ultra-reliable low-latency communication (URLLC), and enhanced mobile broadband (eMBB). Quality of service requirements as well as service parameters within these use cases vary significantly, e. g., URLLC is characterized by an ultralow latency of up to 1 ms, while eMBB has an ultrahigh data transfer rate. The task is to organize the joint provision of such services. The paper proposes a scheme for joint URLLC and eMBB traffic transmission. It is based on the fact that URLLC data has a small volume and can occupy less than one physical resource block. The authors analyze the scheme using the developed retrial queuing system with two orbits - one for temporary storage of interrupted eMBB users and the other for eMBB users waiting for service to start. The authors propose the matrix geometric algorithm for calculating the probability distribution as well as formulas for probabilistic characteristics.

Keywords: 5G; ultra-reliable low latency communication (URLLC); enhanced mobile broadband (eMBB); retrial queuing system; interruption

STATIONARY CHARACTERISTICS OF DISCRETE-TIME Geo/G/1/infinity QUEUE WITH BATCH ARRIVALS AND ONE QUEUE SKIPPING POLICY
  • S. I. Matyushenko  Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation
  • R. V. Razumchik  Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: Consideration is given to the discrete-time single-server system with one queue of infinite capacity and the geometric (Bernoulli) input flow. Customers are homogeneous, arrive in batches, and are served one by one in FIFO (first in, first out) manner. The sizes of arriving batches as well as the service times are assumed to be independent and identically distributed random variables with arbitrary discrete distributions. The queue skipping policy is implemented in the system: upon arrival of a batch, its size is compared with the current total number of customers in the system. If the size of the batch is larger than the system content, all customers residing in the system (including the one in server) are lost and the arrived batch enters the system; otherwise, the new batch leaves the system having no effect on it. Main stationary system performance characteristics, including those of the flow of lost customers, are obtained.

Keywords: discrete-time; queueing system; batch arrivals; queue skipping policy

ON THE DISTRIBUTION OF THE RATIO OF THE SUM OF SAMPLE ELEMENTS EXCEEDING A THRESHOLD TO THE TOTAL SUM OF SAMPLE ELEMENTS. II
  • V. Yu. Korolev  Faculty of Computational Mathematics and Cybernetics, Lomonosov Moscow State University, GSP-1, Leninskie Gory, Moscow 119991, Russian Federation, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: The problem of description of the distribution of the ratio of the sum of sample elements exceeding a threshold to the total sum of sample elements is considered. Unlike other versions of this problem in which the number of summed extreme order statistics and the threshold are fixed, here the specified threshold can be exceeded by an unpredictable number of sample elements. The situation is considered where the threshold infinitely increases as the sample size grows. It is demonstrated that in this case, the distribution of the ratio mentioned above can be approximated by the compound Poisson distribution in which the compounding law is the generalized Pareto distribution.

Keywords: sum of independent random variables; random sum; binomial distribution; Poisson approximation; extreme order statistic; Balkema-De Haan - Pickands theorem; generalized Pareto distribution; compound Poisson distribution

ON MARKOVIAN AND RATIONAL ARRIVAL PROCESSES. II
  • V. A. Naumov  Service Innovation Research Institute, 8A Annankatu, Helsinki 00120, Finland
  • К. Е. Samouylov  Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation , Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: This article is the second part of the review carried out within the framework of the RFBR project No. 19-17-50126. The purpose of this review is to get the interested readers familiar with the basics of the theory of Markovian arrival processes to facilitate the application of these models in practice and, if necessary, to study them in detail. In the first part of the review, the properties of the general Markovian arrival processes are presented and their relationship with Markov additive processes and Markov renewal processes is shown. In the second part of the review, the important for applications subclasses of Markovian arrival processes, i. e., simple and batch arrival processes of homogeneous and heterogeneous arrivals, are considered. It is shown how the properties of Markovian arrival processes are associated with the product form of stationary distributions of Markov systems. In conclusion, matrix-exponential distributions and rational arrival processes are discussed that expand the capabilities of Markovian arrival processes for modeling complex systems, while preserving the convenience of analyzing them using computations.

Keywords: Markov chain; Markovian arrival process; Markov additive process; MAP; MArP

DETERMINISTIC AND RANDOMIZED METHODS OF ENTROPY PROJECTION FOR DIMENSIONALITY REDUCTION PROBLEMS
  • Y. S. Popkov  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • A. Y. Popkov  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation, V. A. Trapeznikov Institute of Control Sciences, Russian Academy of Sciences, 65 Profsoyuznaya Str., Moscow 117997, Russian Federation , ORT Braude College, Karmiel 2161002, Israel
  • Y. A. Dubnov  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation , National Research University Higher School of Economics, 20 Myasnitskaya Str., Moscow 101000, Russian Federation

Abstract: The work is devoted to development of methods for deterministic and randomized projection aimed at dimensionality reduction problems. In the deterministic case, the authors develop the parallel reduction procedure minimizing Kullback-Leibler cross-entropy target to condition on information capacity based on the gradient projection method. In the randomized case, the authors solve the problem of reduction of feature space. The idea of application of projection procedures for reduction of data matrix is implemented in the proposed method of randomized entropy projection where the authors use the principle of keeping average distances between high- and low-dimensional points in the corresponding spaces. The problem leads to searching of a probability distribution maximizing Fermi entropy target to average distance between points.

Keywords: dimensionality reduction; Kullback-Leibler cross-entropy; entropy

DEEP LEARNING NEURAL NETWORK STRUCTURE OPTIMIZATION
  • M. S. Potanin  Moscow Institute of Physics and Technology, 9 Institutskiy Per., Dolgoprudny, Moscow Region 141700, Russian Federation
  • K. O. Vayser  Moscow Institute of Physics and Technology, 9 Institutskiy Per., Dolgoprudny, Moscow Region 141700, Russian Federation
  • V. A. Zholobov  Moscow Institute of Physics and Technology, 9 Institutskiy Per., Dolgoprudny, Moscow Region 141700, Russian Federation
  • V. V. Strijov  Moscow Institute of Physics and Technology, 9 Institutskiy Per., Dolgoprudny, Moscow Region 141700, Russian Federation, A. A. Dorodnicyn Computing Center, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 40 Vavilov Str., Moscow 119333, Russian Federation

Abstract: The paper investigates the optimal model structure selection problem. The model is a superposition of generalized linear models. Its elements are linear regression, logistic regression, principal components analysis, autoencoder and neural network. The model structure refers to values of structural parameters that determine the form of final superposition. This paper analyzes the model structure selection method and investigates dependence of accuracy, complexity and stability of the model on it. The paper proposes an algorithm for selection of the neural network optimal structure. The proposed method was tested on real and synthetic data. The experiment resulted in significant structural complexity reduction of the model while maintaining accuracy of approximation.

Keywords: model selection; linear models; autoencoders; neural networks; structure; genetic alghorithm

IMPROVEMENT OF SELF-TIMED CIRCUIT SOFT ERROR TOLERANCE
  • I. A. Sokolov  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • Yu. A. Stepchenkov  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • Yu. G. Diachenko  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • Yu. V. Rogdestvenski  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: The paper considers a tolerance of self-timed (ST) circuits fabricated with complementary metal-oxide- semiconductor (CMOS) process to short-term soft errors generated by external causes, namely, nuclear particles, cosmic rays, electromagnetic pulses, and noises. Pipeline implementation is usual for practical ST-circuits. Its control bases on handshake between pipeline stages and two-phase operation discipline with a sequence of the working phase and spacer one. Combinational part of the pipeline stage uses dual-rail information signal coding with a spacer. The pipeline stage indication part acknowledges a switching completion of all stage cells, fired at the current operation phase, and generates handshake signals in ST-pipeline stages control. The paper discusses the physical causes of the short-term soft errors. It analyzes soft error types that may appear in CMOS ST-circuits fabricated with 65-nanometer and below standard bulk process. The tolerance level of the proposed soft error hardened ST-register bits is discussed and compared. The paper suggests circuitry and layout techniques improving ST-pipeline soft error tolerance and estimates soft error immunity level for all pipeline parts depending on soft error location.

Keywords: self-timed circuit; tolerance; pipeline; working phase; spacer

EXTRACTION OF CONFIDENTIALITY MARKERS FROM TEXTS UNDER CONDITIONS OF HIGH UNCERTAINTY IN SYSTEMS WITH DATA INTENSIVE USAGE
  • V. I. Budzko  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • V. V. Yadrintsev  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation, Peoples' Friendship University of Russia (RUDN University), 6 Miklukho-Maklaya Str., Moscow 117198, Russian Federation
  • I. V. Sochenkov  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • V. I. Korolev  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • V. G. Belenkov  Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: The main tasks, the results of the solution of which are reflected in the article, are associated with the formation of confidentiality markers when they are used in data-intensive systems under conditions when the composition and structure of the protected information cannot be determined in advance due to the lack of data or the high dynamics of their change, or their definition is not advisable due to the large number of entities whose information is subject to protection. In this paper, an approach is proposed for the formation of confidentiality markers for text materials in the indicated conditions. The article presents the semantic text analysis, which forms confidentiality markers when used to ensure information security in data-intensive systems under high uncertainty in the composition and structure of protected information. The obtained experimental results show that practical implementation of the considered approach in data-intensive systems is promising.

Keywords: confidentiality marker; information security; data-intensive domains; topical cluster; semantics; data leak prevention; intelligent security tasks; text classification; detection of text reuse

CONFLICT VISUAL REPRESENTATION METHOD IN HYBRID INTELLIGENT MULTIAGENT SYSTEMS
  • S. B. Rumovskaya  Kaliningrad Branch of the Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 5 Gostinaya Str., Kaliningrad 236000, Russian Federation
  • I. A. Kirikov  Kaliningrad Branch of the Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 5 Gostinaya Str., Kaliningrad 236000, Russian Federation

Abstract: Small collectives of experts, including specialists from different fields, effectively solve complex problems by analyzing them from different points of view and obtaining a better-integrated solution. A conflict in small collectives of experts can both lead to a deadlock in the decision-making process and generate positive changes: development of the group, diagnostics of relations, and consolidation of the group. A conflict breeds debate, the depth of which allows for more thoughtful and coordinated solutions. Such collectives of experts solve problems effectively. Thus, modeling of their work and possible conflict situation with managing it allows developing a decision-support method that is relevant to solving a complex problem. Visualization of a conflict situation makes appeared contradictions contrast and observable. In the research, the authors represent a collective of agents-experts in the form of an undirected weighted graph. The methods of graph visualization are considered. To visualize problem- and process-oriented conflicts within hybrid intelligent multiagent systems, the authors propose a method based on the spring model of graph drawing.

Keywords: collective of experts; conflict; visualization of conflict

ESTIMATING THE FAIR VALUE OF OPTIONS BASED ON ARIMA-GARCH MODELS WITH ERRORS DISTRIBUTED ACCORDING TO THE JOHNSON'S SU LAW
  • A. R. Danilishin  Department of Operations Research, Faculty of Computational Mathematics and Cybernetics, M. V. Lomonosov Moscow State University, 1-52 Leninskie Gory, Moscow 119991, GSP-1, Russian Federation
  • D. Yu. Golembiovsky  Department of Operations Research, Faculty of Computational Mathematics and Cybernetics, M. V. Lomonosov Moscow State University, 1-52 Leninskie Gory, Moscow 119991, GSP-1, Russian Federation, Department of Banking, Sinergy University, 80-G Leningradskiy Prosp., Moscow 125190, Russian Federation

Abstract: In continuation of the article " Risk-neutral dynamics for the ARIMA-GARCH (autoregressive integrated moving average - generalized autoregressive conditional heteroskedasticity) random process with errors distributed according to the Johnson's SU law," this paper presents the experimental results for the ARIMA-GARCH (autoregressive integrated moving average - generalized autoregressive conditional heteroskedasticity) models with normal (N), exponential beta ofthe second type (EGB2), and SU Johnson (JSU) error distributions. The fair value of European options is estimated by the Monte-Carlo method based on the results obtained in the specified article by using the extended Girsanov principle. The parameters of the ARIMA-GARCH-N, ARIMA-GARCH-EGB2, and ARIMA-GARCH-JSU models were found by the quasi-maximum likelihood method. The efficiency of the resulting risk-neutral models was studied using the example of European exchange-traded options PUT and CALL on basic assets DAX and Light Sweet Crude Oil.

Keywords: ARIMA; GARCH; risk-neutral measure; Girsanov extended principle; Johnson's SU distribution; option pricing

APPLICATION OF MULTISCALE APPROACH AND DATA SCIENCES FOR MODELING THERMAL CONDUCTIVITY IN LAYERED STRUCTURES
  • K. K. Abgaryan  A. A. Dorodnicyn Computing Center, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 40 Vavilov Str., Moscow 119333, Russian Federation, Moscow Aviation Institute (National Research University), 4 Volokolamskoe Shosse, Moscow 125080, Russian Federation
  • I. S. Kolbin  A. A. Dorodnicyn Computing Center, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 40 Vavilov Str., Moscow 119333, Russian Federation, Moscow Aviation Institute (National Research University), 4 Volokolamskoe Shosse, Moscow 125080, Russian Federation

Abstract: Modeling thermal properties of layered structures is currently a popular area of scientific research. This is due to the constantly growing speed of operation of microelectronic elements often based on layered structures that release more and more energy during operation in the form of heat which must be removed to avoid overheating and loss of functional properties of devices. The paper presents an integration approach that allows one to combine the methods of multiscale modeling and data analysis. It is shown that application of this approach makes it possible to obtain a new quality when solving the problem of constructing a model of heat transfer in a two-layer GaAs/AlAs structure. The effectiveness of use of machine learning methods for analyzing the dependence of the effective thermal conductivity coefficient of laminated materials on structural features and external factors is shown. The development of the proposed approach will be able to provide formation of information for reasonable selection of materials for layered structures for microelectronic devices.

Keywords: multiscale modeling; integration approach; layered structures; predictive modeling; kinetic Boltzmann equation, thermal conductivity coefficient; data analysis methods

ABOUT DIGITAL LITERACY AND ENVIRONMENTS FOR ITS DEVELOPMENT
  • V. B. Betelin  Federal State Research Center Scientific Research Institute of System Development, Russian Academy of Sciences, 36-1 Nakhimovsky Prosp., Moscow 117218, Russian Federation
  • A. G. Kushnirenko  Federal State Research Center Scientific Research Institute of System Development, Russian Academy of Sciences, 36-1 Nakhimovsky Prosp., Moscow 117218, Russian Federation
  • A. L. Semenov  M. V. Lomonosov Moscow State University, 1 Leninskie Gory, GSP-1, Moscow 119991, Russian Federation , Alex Berg Institute of Cybernetics and Educational Computing of Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation , Moscow Institute of Physics and Technology (National Research University), 9 Institutskiy Per., Dolgoprudny, Moscow 141701, Russian Federation
  • S. F. Soprunov  Center for Pedagogical Excellence, 15-1 Bol'shaya Spasskaya Str., Moscow 129090, Russian Federation

Abstract: Digital literacy is becoming a key characteristic of a XXI century person. It can and should be formed from an early age. Digital environments can be environments for achieving subject, metasubject, and personal educational outcomes. The work analyzes domestic and international experience in this direction, beginning in 1960s, offers a general view of the "big ideas" mastered in digital environments and provides a system of principles that ensure the effectiveness for educational purposes of digital environments and systems developed under the guidance with the participation of the authors, they are effectively used in the Russian Federation in the formation of digital literacy in young children, the basis for computational thinking _ systemic thinking of the XXI century. The article gives an overview of the results obtained in this direction over the last years. Special attention is paid to the problem of visualization and representation in the real world of algorithmic processes and the programs that set them.

Keywords: informatics; computer science; digital literacy; computational thinking; robot; programming language; Logo; PervoLogo; LogoFirst; PictoMir; learning environment; visualization

EVOLUTION OF CLASSIFICATIONS IN SUPRACORPORA DATABASES
  • A. A. Goncharov  Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • I. M. Zatsman  Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation
  • M. G. Kruzhkov  Institute of Informatics Problems, Federal Research Center "Computer Science and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow 119333, Russian Federation

Abstract: The paper examines the task of recording changes to descriptions of meanings of German modal verbs in the process of annotating parallel German-Russian texts within a supracorpora database. This task was used as a case study to analyze the specifics of using dynamic classification systems (DCS) in information systems. The distinctive feature of a DCS is that semantic content of its concepts may change in the process of annotation which often entails the need to reclassify previously annotated data according to the changes made. This paper aims to answer the following questions: (i) What factors may have an impact on the need to edit and/or reclassify the annotations created prior to the concept changes? and (ii) What kind of operations may be used to represent the changes to concepts in the DCS? The paper describes seven types of possible changes and enumerates the corresponding operations applied to the DCS concepts in the process of annotation. The operations are grouped in three categories depending on how they affect the need to reclassify the previously created annotations.

Keywords: dynamic classification; faceted classification; reclassification; supracorpora databases



 

RUS