Інформаційна система виявлення прихованого змісту текстових повідомлень. Частина 3. Розробка програмно-технологічного забезпечення інформаційної системи виявлення прихованого змісту в текстових повідомленнях

Abstract

This work consists of four chapters. In first chapter an analysis of the principles of working with the Big Data was carried out, the existing approaches to the analysis of text information and methods of work with it were examined. As a result, conclusions are drawn about the possibility and feasibility of using principles of work with the Big Data, as well as using approaches to the analysis of natural language to reveal hidden content in it. In second chapter is described the design process of an information system that can analyze textual information and make considerations about the data that it analyzes. It is possible because of performing natural language processing and text tone analysis and combining their results together. Joint result with high reliability will represent real entities and their relationships in text together with the author’s attitude towards them. In third chapter is described the development process of an information system that can analyze textual information obtained by recognizing the natural language using IBM Cloud services and transfer it to Azure DevOps cloud simplifying routine management processes. As a core of this information system stands Python 3 script that designed for Jupyter Notebook and divided into sections that responsible for different processes like linking with Azure DevOps Cloud and IBM Cloud, input file reading, API calls for data analysis using text tone recognition and natural language understanding services. In fourth chapter information system for hidden content detection in text information results brief description and analysis are made. Also, an experiment for comparison and texting of developed Information System was made. From results of this experiment becomes clear that for the same situations developed IS works same or, sometimes, better than human expert for much less time. Depending from the text size, difference can be up to thousands of times or more. This key feature allows to implement near-to-real-time functionality that detects text topics easily.