An approach to the construction of an adaptive architecture for enterprise software has been developed in order to increase the efficiency of automated processing of documents using semantic and cognitive technologies. The proposed approach takes advantage of the existing methods of organizing the architecture of applied software. It is substantiated that the architecture of the developed software for automatic cataloging should have a hybrid clientserver implementation, including elements of modular and microservice architecture. It is shown that a significant reduction in the costs of cataloging, checking the completeness and inventory of documentation, as well as an increase in the quality of design are provided through the semantic analysis of documentation using a knowledge base that is updated automatically.
3. Ability to localize the location of the error, which allows, with a good organization of modules, to correct defects in one module, causes errors in another module. 4. Fast recompilation while fixing the error. 5. Ability to reuse modules. 6. Tools are provided to solve each processing task. 7. High resiliency due to redundancy of critical services.
Availability of analytics tools, so it is easy to track dependencies between services [4]. The developed architecture includes elements from the hybrid architecture (Figure 1): 1. Element "Client". 2. Element "Application Server". 3. Element "Data warehouse". 4. Element "Complex of microservices".
Application Server Container
Processing module
Services
2. Development of modules for hybrid client‐server implementation
The client module interacts with personnel and provides data in various formats: 1. Tabular. 2. Graphic. 3. Graph. 4. Text [5].
To solve visualization tasks, built-in components for working with knowledge graphs are used, data is provided in each of the required formats with the ability to navigate the knowledge graph. Also, to solve analytic tasks, components for working with multidimensional data are used. The graphical interface is also designed for automated structuring of knowledge with the participation of subject matter experts [6]. 2.2.
The application server is built on a modular basis and includes: 1. The core. 2. Modules of data analysis. 3. Modules of data processing.
4. Modules of data fusion. 2.3.
The flexible core is the central component of the software being developed; it interacts with the rest of the modules and processes user requests from the graphical interface. The kernel, in the process of functioning, processes user requests, interacts with data stores and provides the user with requested samples or calls processing functions from plug-ins [7, 8]. 2.4.
Modules interact with external and internal services that implement various stages of working with data. The search for the required service is carried out by the service register (a software module that interacts with the ontological description of the service model), which is associated with the service model of the ontology. 2.5.
The data processing module includes three stages.
Stage "Data preprocessing".
Preprocessing is aimed at noise reduction in order to improve visual perception for subsequent data processing. For some elements (in particular, for numerical data), the preprocessing stage is skipped (Figure 2)
Pre-processing algorithms
Raw data in text format
Noise-free text in string format Dictionaries, templates
This module implements data fusion. Merge refers to the process of data binding in accordance with an ontological model in order to ensure the integrity and consistency of data. The data integration diagram is shown in Figure 5. Integration
Related data Data in the format of
triplets "subject" "predicate" - "object" 3. Development of storage modules, diagnostics and microservice architecture 3.1. Description of the data warehouse element and the data storage object model
3.2.
The role model (Figure 6) describes the roles involved in the processes. The role model includes the staff of the institution, subdivided into laboratory assistants, management personnel, and engineering workers.
Actor
Staff Full name
Branch
Diagnostics Date
Executor Engineering and technical
personnel
Qualification Technician
Engineer Manufacturing facility
Manufacturing facility Figure 6: Object model of institutional personnel Managing staff
Laboratory assistant Director
Foreman
Analysis Laboratory assistant Measuring control Measurement Figure 7: Object model of diagnostic tools 3.3.
Data on the documents collected in the institution's IS are represented by text records and numerical values. Numerical and qualitative data are highlighted. The data type hierarchy includes subjective and objective data. Each type has a qualifying field.
The company's activities are carried out within the divisions of the institution. In addition to the production process carried out in the workshops, the laboratories carry out analyzes of the manufactured products. The model is shown in Figure 9.
Subdivision
Manager Manufacturing facility
Laboratory
Management (department)
The source of algorithms in the technological process of processing are services, which are accessed through the corresponding modules. Services are used both in the process of preprocessing data and in solving problems received from users. The results of processing on demand from users are also saved to the knowledge base in order to speed up the execution of similar tasks in the future. Thus, despite the close connection between the modules, their independence is preserved, and the module itself remains operational, provided the kernel and data stores are preserved [15, 16].
The modules considered earlier include a set of components intended for processing data within a module and interacting with each other through interfaces. Internal kernel modules interact by calling the API methods of the components, the interaction of the graphical interface with the kernel, and the kernel with the data source is carried out by sending GET RESTAPI requests or through the Web-socket (Figure 10).
"Customer" Components for generating queries and displaying processing results
Text preprocessing
Preparing the text
Server
Figure 10: Modules of the developed software
The paper was prepared with the financial support of the Ministry of science and higher education of the Russian Federation in the course of the applied research «The comprehensive project to create high-tech production of software tools for automatic analysis of documentation on paper and digital media using semantic-cognitive technologies for cataloging poorly structured information» (unique identifier of the project 075-11-2019-055, Decree of the Government of the Russian Federation N 218, 09.04.2010).
[1] Saurabh Gupta and Anil Kumar Meena, A Practical Implementation of Automatic Document Analysis and Verification using Tesseract, International Conference on Computational Techniques, Electronics and Mechanical Systems (CTEMS), December 2018. doi:10.1109/CTEMS.2018.8769310. [2] S. Marinai, Introduction to Document Analysis and Recognition, in: S. Marinai, H. Fujisawa (eds), Machine Learning in Document Analysis and Recognition. Studies in Computational Intelligence, Springer, Berlin, Heidelberg, 2008, vol. 90, pp. 1-20. doi:10.1007/978-3-540-76280-5_1.