Search

This paper presents new methods for entity recognition and relation extraction tasks on partially labeled and unlabeled datasets. The proposed methods are based on techniques of semi-supervised, unsupervised and the transfer learning. We use the few-shot learning technique to construct specific algorithms for the new data sources without manual retraining. To compare the results with other studies, we conducted experiments on two benchmark datasets for the Russian language. The results for named entity recognition demonstrate significant improvement and outperform the state-of-the-art results. Our results for relation extraction are comparable to other research. We assume that a longer BERT fine-tuning will help to improve them, and we also plan to experiment with other few-shot learning methods in the near future.

The Russian language has an inflective structure and does not have a strict word order. This causes processing difficulties, such as part-of-speech homonymy. This article is devoted to the mentioned issue. The existing approaches to resolving the morphological homonymy problem can be divided into the following groups: rule-based approaches, statistical approaches, machine learning approaches, and combined methods. In the paper, we showed that each approach has its advantages and disadvantages; however, combining several approaches can significantly increase the precision of the algorithm. Moreover, the article provides the analysis of the influence of certain features on the morphological homonymy resolution. The precision of the proposed algorithm is sufficient for its use in the tasks of intellectual text processing texts, for example, in machine translation and summarization systems. The proposed method is successfully used in the geographic location system. The main problem is the distinction between function words (conjunctions, particles, prepositions, interjections). Solving this problem is one of the priorities for the further work. We also plan to implement a system without a dictionary, in order to determine better morphological features for unknown words.

The paper is dedicated to applying a hybrid approach based on rules and machine learning for anaphora resolution in the Russian language. The model combines formal rules, the Extra Trees machine learning algorithm and the Balance Cascade algorithm for working with imbalanced learning sets. A number of features were obtained from the rules or were generated from other features; in addition, the syntactic context was taken into account. A neural network algorithm SyntaxNet was used to analyze the syntactic context.

Analysis of experimental results of the binary noise-immune coding with one-dimensional modulation best achievements points out the impossibility of achievement the continuous Gaussian channel parameters of such scheme in practice. In the paper is exhibited that changeover to more effective two-dimensional non-binary PAM modulation with ternary noise-immune codes in each quadrature branch of a modem allows to work with high enough reliability and to reach frequency-energetical system parameters that are equal to analogous parameters of continuous channel with AWGN. Symbol error probability Q s =10−4 for E b /N 0 =4.963 and total spectral efficiency 4.036 bit/s/Hz have achieved for best modulation and coding scheme from proposed set.

The main problem considered in this paper is creating algorithms for estimation the relevance of documents to the search query on the basis of sentences structure analysis. To decide this problem, we use the relations between words constructed by the program system Link Grammar Parser, based on the so-called link grammar. There were suggested the natural system of links for Turkic languages, created prototypes of Link Grammar Parser for Kazakh and Turkish. In addition, we considered some generalization of the summarization algorithm of Niraj Kumar. Results of research are planned to use in different information retrieval systems.

Compound methods for solution of boundary problems for irregular waveguides with the use of statements of carry and averaging are suggested.

The method of even sampling for solution of ordinary differential equation is proposed. Its high effectiveness is demonstrated on the example of propagation of H 01 -mode in an irregular waveguide with a circular section.

Potentially effective schemes of frequency doublers with cross-section modulation of electron beam are considered.

this article proposes the complex analysis of quality assuring system of distance education, and analyzes the experience in distance education at the BSUIR and other universities of Belarus.

Considered in this paper are the features of Cherenkov oscillators simulation. The variants for electron beam accelerating voltage 250kV and a current 1-2kA are discussed. Using of ultrahigh power microwave impulses Pu=1-3 GW with duration tauu=0.5-5 mus within lambda=1-10 cm band makes it possible to implement effects electromagnetic defeat. Long-term experimental and theoretical research of different types of microwave oscillators has shown, that in order to achieve specified radiation parameters the Cherenkov oscillator should be used. Presented are the models and efficient techniques, which allow to carry out complete optimization of electrodynamic system profile (calculation takes about 30 seconds). That would make for discovery of new design types. Considered in this paper are simulation features and the new oscillators for electronic beam with the given accelerating voltage 250kV and current 1-2kA are discussed.

Correct procedures of division of an electromagnetic field excited by an electronic beam in electrodynamic system of a TWT on solenoidal and potential components on backward and forward waves are proved. It is specified on this division incorrectness in the simplified "one-wave" theory of the TWT resulting as to numerical, and physical mistakes.

The new sight at the mechanism of distant propagation of EMF-waves in alive biological objects has been developed. This mechanism is determined by arising of dynamics negative conduction of membrane of cell in the presence of transportation through it positive and negative ions K*, Na*, Cl/sup -/, NO/sub 3//sup -/, and brighten up of membrane at high frequency, and re-radiation the waves by the acoustic-electrical oscillations of membrane.

Presented in the paper is two-dimensional theory of relativistic TWT-BWO taking into account cut-off modes. Numerical realization of the theory is based on application of a special method of block matrix prorace for decision of a point-to-point boundary problem for ODE of amplitudes of the connected waves. Optimum variants of the relativistic generator are found and investigated. Attention is paid to absence of conditions of periodicity of the second type (Flocke conditions) in a field of the matched section of periodic irregular waveguide.

Results of developing a nonlinear theory of relativistic TWTs and BWT-Os built around irregular hollow waveguides with regard to both propagating and beyond-cutoff waves are given. Robust computational procedures with allowance made for beyond-cutoff waves are covered in great detail. The results are given of optimizing various relativistic TWT-Os by efficiency.

The mass distribution of particles produced in the high-current vacuum arc was investigated. The experiments were concentrated on evaluating the spatial mass distribution emitted in radial as well as in azimuthal directions calculated from mass deposition profile on the collectors surrounding the arc discharge. The experiments were carried out in a vacuum chamber evacuated to ambient pressure

The paper presents optimizing high voltage stand-off insulators for vacuum devices which work at alternating voltage at power-line frequency. In order to achieve the maximum surface electrical strength of the insulator, optimizing calculations were made concerning both the shape of the side surface of insulating material and the shape of electrodes controlling distribution of the electric field (insulator's ferrules). The article presents results of optimization and experimental insulator research.

The paper presents results of studies on switching unloaded power cable and physical models of medium voltage cable lines by a vacuum switch. The performed experiments have shown the occurrence of early and late arc reignitions, generation of significant overvoltages and overcurrents as well as their dependence on the length of switched unloaded power cable, parameters of the source side circuit and vacuum switch.