AUTOMATIC DOCUMENTATION AND MATHEMATICAL LINGUISTICS最新文献

The paper describes three stages in the construction of a fuzzy classifier. The first refers to the formation of fuzzy rules, the second stage is feature selection, and the third stage is optimization of membership functions parameters. The influence of clustering methods on the efficiency of the formed fuzzy classifier rules was estimated by three different fitness functions. These functions were total variance, the Davies–Bouldin index, and the Calinski–Harabasz index. The grasshopper optimization algorithm was binarized using S- and V-shaped transformation functions for feature selection. The constructed classifiers have been tested on datasets from the KEEL repository.

This article defines the set of inference rules for JSM reasoning. Hypotheses concerning the causes generated by these rules form empirical regularities (ERs). Many types of ER (the intension of the concept of ERs) form a lattice. Types of ERs can be represented by trees of a special type.

The disadvantages of modern methods of classifying objects and processes are considered. We propose a method of constructing three-dimensional classifications that allows the elimination of some of these shortcomings, based on the system-object approach, using the ideas of multidimensional classification and natural classification. Three basic system characteristics are used as classification planes: structural (node), functional (function), and substantive (object), which allows for the classification by types of functional request to the system from a higher-order system (supersystem) by types of system formation processes and by the obtained results. Each classification is a tree-type graph with one vertex that is common to all three planes. The formal description of the three-dimensional graph by means of descriptive logic is presented, which not only allows for the phenomena and objects of the subject area to be classified, but also for the cause-and-effect relations existing in this area to be traced. The procedures for the use of three-dimensional system-object classification for forecasting and management support are described. An example of three-dimensional classification for functional diagnostic devices is given.

A study has been conducted to determine the feasibility of referring patients with chronic pancreatitis to computed tomography using the JSM method of automated research support. This study structurally repeats a similar study from 2022 but uses refined heuristics for finding empirical regularities, while for the first time using the union of predictions for various strategies of the JSM method, provided that the predictions are consistent. A comparison of the results of the described study and the study of 2022 is provided.

An alternative approach to speech denoising using generative diffusion models that model the distribution of training data is proposed. In recent years, such models have led to promising results to be obtained in the field of generating signals of various kinds, and these are superior in many ways to previous generative models, such as variational autoencoders. However, diffusion models have not yet found wide application in the field of speech denoising. A new diffusion model is presented, which can be used to denoise real speech signals using a deep neural network. Our own data set, with more than 150 h of pure speech in Russian, has been created. The obtained results, estimated using the metrics scale invariant signal to distortion ratio and perceptual evaluation of speech quality, are comparable or superior to the results of the best discriminative models.

To analyze the functioning of servers in a special-purpose information system in different modes of operation, mathematical models are proposed, the formation of which is based on the methods of the theory of mass service systems. These models allow the functioning of the system servers to be analyzed in two modes of request processing—synchronous and asynchronous—and estimating the waiting time of requests that are in the queue.

We conducted an investigation into the patterns of international collaboration among publications by Russian scholars, specifically articles and reviews, with their counterparts from Iran, India, and Turkey during the period spanning from 2011 to 2021. The selection of these countries was based on their growing role on the world science stage and increased partnership with Russia over the last ten years. The primary objectives of this study are to conduct a comprehensive analysis investigating the patterns of Russian collaborations with selected countries. This includes revealing the disciplinary priorities in their international collaborations, identifying the top organizations involved, and understanding the model of co-authorship. The findings indicate a growing partnership across various disciplines, with a particular focus on basic research in physics and clinical medicine. The top collaborating organizations in Russia, as well as in each of the studied countries (Iran, India, and Turkey), are either wholly or partially funded by their respective national governments. Furthermore, the values of bibliometric indicators for these collaborative publications indicate a high level of quality. Our findings could benefit decision-makers by providing evidence-based data and funding agencies.

An analysis of methods for evaluating the effectiveness of complex system design processes is presented, and formal indicators for assessing their quality based on the mathematical apparatus of fuzzy sets are proposed. A complex system, such as a model of the business processes of designing, testing, and manufacturing (creating) on-board radio-electronic equipment was chosen as the basis, involving multiple departments of the enterprise, from managers, circuit designers, and designers to the procurement department, testing workshops, and support services. It is proposed to consider all components of business processes as resources, to identify material, labor, financial, temporal, hardware, and software resources among them, establish a connection between them, and evaluate the weight coefficients of the connections to find the most effective implementation of the complex system design process using a radar diagram of relative quality indicators. An E-network logical-dynamic model of project design work execution is used to evaluate the efficiency of each resource over time and calculate the integrated efficiency indicator of the life cycle business processes of creating a complex system. This model allows for the consideration of dynamic characteristics of parallel interacting departments and project stages. Based on the fuzzy set apparatus, an approach to evaluating the effectiveness of each stage of the complex system creation project is considered, which in turn allows for the uncertainty and fuzziness in evaluation to be taken into account and determine the importance of each criterion based on investor preferences. This helps make informed decisions about building an investment portfolio and achieving desired financial goals.

The results of a study of high-frequency key terms in the subject of SciVal (Scopus) are presented, focusing on the term “bibliometric analysis.” The array of high-frequency key terms collected for May to November 2022 from 181 SciVal topics, in which the term “Bibliometric Analysis” appeared as the high-frequency key term, was analyzed by three approximately equal groups of co-words, formed by the number of total intersections in the topics. This division of the high-frequency key terms fits well into S. Bradford’s law, due to which the “core” of the high-frequency key terms was formed on the topics under study. As a result, the topics closest in content were identified. The results of our study form a contribution to the understanding of the relationships between different research topics by analyzing the dynamics of keywords, confirming the hypothesis that using networks of keywords from different disciplines, you can identify common features between them and the number of matches between keywords affects the strength of relationships between topics.

This paper considers the problem of the generation of descriptors to reduce data volumes, text data resources, and search times through the use of the new factors of authorship, region, emotive meaning, and popularity, as well as a text category without special marks that can be used to generate descriptors. This approach allows the use of unique lexical-grammatical and distributive text regularities. The research outcomes can be used to determine authorship and text type.