Software & Systems

The paper describes the current state in the field of automatic term extraction from specialized natural language texts, including scientific and technical documents. Practical applications of methods and tools for extracting terms from texts include creation of terminological dictionaries, thesauri, and glossaries of problem oriented domains, as well as extraction of keywords and construction of subject indexes for specialized documents. The paper overviews approaches to automatic recognition and extraction of terminological words and phrases, which cover traditional statistical methods and methods based on machine learning by using term features or with modern neural network transformer-based language models. A comparison of the approaches is presented, including quality assessments for term recognition and term extraction. The most well-known software tools for automating term extraction within the statistical approach and learning by features are indicated. Authors' studies on term recognition based on neural network language models are described, being applied to Russian scientific texts on mathematics and programming. The data set with terminological annotations created for training term recognition models is briefly characterized, the dataset covers the data from seven related domains. The term recognition models were developed on the basis of pre-trained neural network model BERT, with its additional training (fine-tuning) in two ways: as binary classifier of candidate terms (previously extracted from texts) and as classifier for sequential labeling words in texts. For the developed models, the quality of term recognition is experimentally evaluated, and a comparison with the statistical approach was carried out. The best quality is demonstrated by binary classification models, significantly surpassing the other considered approaches. The experiments also show the applicability of the trained models to texts in close scientific domains.

The paper discusses the problem of planning flow-type production processes. In terms of a cascade scheme, the complex solution covers the stage of assigning preparatory units and the subsequent stage of forming detailed technological routes to fulfill a given set of requirements on time and taking into account the constraints on permissible processing durations at each processing stage. This scheme comes as a part of a problem-oriented computing complex. However, due to a number of natural reasons, the problem may become inconsistent right at the stage of assigning preparatory units. One of the ways to overcome these difficulties is to develop and implement penalty function algorithms to find the maximum joint subsystems in inconsistent optimization problems. The paper proposes an ideologically different approach for this purpose. It is based on considering the preliminary stage of requirement placement in such a way that the subsequent stages of problem-solving process are guaranteed to be solvable. The requirement placement is formalized as a search for an optimal mapping that minimizes “potential” workload on preparatory units during the planning period. To solve this problem, the authors of the paper have developed a genetic algorithm, which resulted in a significant advantage in terms of speed in comparison with fundamental approaches of mathematical programming (for example, integer linear programming models). In order to reduce the risk of population extinction at each iteration of the genetic algorithm, the authors apply the rule of unconditional migration of a representative with the lowest criterion value. This approach also provides effective convergence indices of the algorithm in terms of the number of iterations without significant improvement of the objective function. The developed genetic algorithm is implemented as a stand-alone module of a computing system for solving process manufacturing scheduling problems. The authors conducted a computational experiment using this module in terms of a comparative analysis of the solution quality of the initial complex problem.

This relevance of the research is due to the reliability of software components of cyber-physical systems being a key component of their effective functioning. Its appropriate mathematical modeling is essential for the economy digitalization. The paper aims to eliminate the disadvantages of known approaches to modeling the reliability of software components, when the estimates of reliability characteristics are based on empirical data on the errors detected during the testing of programs. Hence, the testing results mostly depend on test duration and on the completeness of the processed data area coverage by the sub-area of the data generated during testing. This reduces the efficiency of reliability estimation. The research focuses on the reliability modeling methods for software components of cyber-physical systems. The reliability is characterized by the delay time in the feedback loop between components. The authors of the paper used methods of software engineering, reliability theory, probability theory and Markov processes. The main result is mathematical reliability models of software components of cyber-physical systems. They combine semi-Markov models of software components, generations of their faults and failures. The developed mathematical models are based on the structural-parametric semi-Markov model of software faults and failures. Its parameters are determined by the computational complexity and requirements to the software taking into account its functional purpose. The authors obtained formalized descriptions of Poisson flows of faults and failures of software components of a cyber-physical system. The practical relevance of the paper is due to its application for determining the reliability of software components at all stages of a cyber-physical system life cycle, where elements interact, self-adjust and adapt to changes using standard software-implemented protocols.

This paper focuses on specification and verification of software systems and their components. It researches a unified specification language that correlates with both random testing systems and static verification tools based on type systems. A variety of programming languages, configuration systems, deployment and other tools require developers to make efforts to integrate them. Verifiable component specifications help to simplify the task. The paper proposes an approach to a unified specification representation integrated with systems for both static type checking and dynamic testing. This solution relies on methods of applicative computing and type theory. It is a conceptual framework for building specifications embedded in various software environments. The lack of static verification capabilities due to limited type systems is eliminated by dynamic testing to some extent. The author implements testing by interpreting specifications into definitions for property-based random testing systems. The practical significance of the proposed approach is automation of the process of constructing typed wrappers, or facades, which are essential for using components from less typed environments in programming languages with more expressive type systems. The approach automates both the verification of such wrappers and the methods of their construction by defining specification refinement operations. In practice, this allows detecting errors in typing of third-party components at early development stages. The paper gives examples of program specifications with side effects. A basis for specifications is category theory formalizations. The author also analyzes approaches to translating specifications into other representations and to iteratively improving specifications by transforming them.

The paper is relevant due to the tasks in the field of information security that require comparison of programs in their different representations, for example, in text assembly code (e.g., for vulnerability search or authorship verification). The paper presents a proximity metric for two texts in the form of a character string list, which is a development of its previous author's version. The main result of the current study (as a part of the main study aimed at genetic de-evolution of programs) is the metric itself, as well as its characteristics and peculiarities revealed through experiments. The paper presents the metric in analytical form implemented in Python. The metric takes at the input two lists of character lines for comparison, and the coefficients of taking into account the element position from the beginning of the list and the character sequence. The calculation result is a numeric value in the range from 0 to 1. Metric's novelty is in a sufficiently accurate and sensitive assessment of two texts' proximity regardless of data representation formats. The current metric version differs from the previous one by taking into account the mentioned coefficients. Theoretical significance lies in the development of comparing methods for arbitrary texts that are a list of character lines containing information, which appears sequentially according to a certain logic (requires position consideration). Besides the general purpose of comparative tools like this, the metric is practically relevant due to the possibility of determining the proximity of two programs. These programs have a binary representation of the machine code. It is pre-transformed into a textual representation of an assembly code.

The paper describes the implementation of a neuro-fuzzy network (NFN) model based on Bayesian logical-probabilistic fuzzy inference model as a specialized software tool. In addition to the NFN model itself, it implements adapted algorithms for constructing (grid partitioning algorithm for generating fuzzy rules) and training (error backpropagation algorithm and hybrid algorithm) NFN known from ANFIS network in MATLAB. The designed software tool is written in Java. It solves practical problems and investigates the effectiveness and applications of NFN based on the Bayesian logical-probabilistic fuzzy inference model. The authors discuss the experience of building and training the proposed NFN type using the developed toolkit. The authors also consider the creation and training examples of NFNs designed to solve specific approximation problems of multivariable functions based on real open and synthetic data sets. The paper compares the results obtained using the developed tool and the ANFIS tool from MATLAB. The authors confirmed that the proposed NFN model can be a universal approximator of complex functional dependencies. This confirms its efficiency and possibilities of using it in different fields. The inclusion of various approximation quality metrics in the program allows evaluating network training quality, accuracy, stability, and adaptability to new data comprehensively. Access restrictions for Russian users to foreign commercial software enhance the practical significance of the developed software tool based on the original NFN model; it makes the development relevant and useful for a wide range of users.

The purpose of this paper is to develop an approach in order for engineering methods of equipment calculation to form as separate programs integrated with simulation-based CAD systems, and the results to be input data for a CAD-system. The research focuses on toluene settling from water-methanol mixture, the particular emphasis is on non-standard technological equipment of the settling process (settler). The authors present technological calculations for material and energy balances in a universal simulation program. They determine the geometric dimensions of non-standard equipment using the developed program. A 3D model of the equipment is formed in a graphical CAD-system. The approach is implemented in the form of a program in VBA language, which is connected with the model of a settling process. VBA language is basic within the software used and widespread in automating calculations within the system. Since the program database does not contain a settler calculation module, the authors have developed this module using the Import User Model tool. A parametric analysis was conducted using the Case Study add-on. The authors identified the maximum allowable methanol content in the initial mixture (up to 0.3 mass fraction), at which the density of the water-methanol mixture exceeds the density of the toluene phase. The authors calculated the main dimensions of the equipment using engineering methods, which are implemented as a separate program linked to the model in the simulation CAD system. The authors have developed a 3D model of the equipment body in a graphical CAD system using variables. It is the basis of the assembly of the settler, supplemented with standard elements from the program database. As a result, the developed program determines settler dimensions; it links modeling and graphical CAD system. It allows automating engineering methods of energy efficient equipment calculation.

The paper considers an integrated nesting and routing task. It combines the problems of nesting optimization and cost minimization of the cutting process with numerical control (CNC). To solve the problem and develop appropriate optimization algorithms, it is necessary to obtain scientifically substantiated data. Specifically, the data on the values of the cost parameters of a sheet cutting process on the CNC technological equipment for different grades and thicknesses of processed materials. The paper describes a database of such cost process parameters. There is also its diagram, basic structure and organization. The authors chose a relational model to store the data on cost parameters. It consists of 8 tables and contains information on grades and thicknesses of materials processed, cost and density values of materials, cost per unit of cutting tool path at working and idle speeds, as well as one piercing point. Usability of working with data in the database (visualization, adding new records, deleting records, calculating cost parameters for new material grades and thicknesses, changing parameter values) is provided by the designed software in Python. The authors propose to use it either as a separate product or together with the software for automatic design of control programs for CNC sheet cutting equipment to solve practical problems. The authors show this work significance on a model example of designing nesting and tool routing for one type of a laser cutting machine.

The research focuses on a system analysis of the subject area, particularly the processes of wastewater and soil microbiological treatment and the database development. The paper is relevant due to the problem of unstructured large volumes of incoming heterogeneous information in this area. The authors analyzed the literature describing the existing databases on microbiological treatment and indicating their advantages and disadvantages. They justified the relevance of developing a database that integrates all key components of microbiological treatment processes of wastewater and soil. The authors conducted a systematic analysis of the subject area. Using a systematic approach, they constructed a data storage architecture. They noted the advantages of the developed system and showed examples of search query execution. The developed the Microbiological Purification database contains extensive information on pollutants and microorganisms with descriptions of microbiological cleaning processes. The proposed data storage system is useful for researchers whose area of scientific interest is microbiological cleaning processes, microbiology, chemistry and chemical technology. It allows reducing the time for information retrieval during scientific research work.