Decompositional Extraction and Retrieval of Conceptual Knowledge

• Дата: 2023
• Автори: Terletskyi, D.O., Yershov, S.V.
• Формат: Стаття
• Мова: Ukrainian
• Опубліковано: Інститут програмних систем НАН України 2023
• Теми: internal semantic dependencies; decomposition consistency; decomposition of classes; knowledge extraction; knowledge retrieval; UDC: 004.82:004.832.2
• Онлайн доступ: https://pp.isofts.kiev.ua/index.php/ojs1/article/view/516
• Назва журналу: Problems in programming

An ability to extract hidden and implicit knowledge, their integration into a knowledge base, and then retrieval of required knowledge items are important features of knowledge processing for many modern knowledge-based systems. However, the complexity of these tasks depends on the size of knowledge sources, which were used for extraction, the size of a knowledge base, which is used for the integration of extracted knowledge, as well as the size of a search space, which is used for the retrieval of required knowledge items. Therefore, in this paper, we analyzed the internal semantic dependencies of homogeneous classes of objects and how they affect the decomposition of such classes. Since all subclasses of a homogeneous class of objects form a complete lattice, we applied the methods of formal concept analysis for the knowledge extraction and retrieval within the corresponding concept lattice. We found that such an approach does not consider internal semantic dependencies within a homogeneous class of objects, consequently, it can cause inference and retrieval of formal concepts, which are semantically inconsistent within a modeled domain. We adapted the algorithm for the decomposition of homogeneous classes of objects, within such knowledge representation model as object-oriented dynamic networks, to perform dynamic knowledge extraction and retrieval, adding additional filtration parameters. As the result, the algorithm extracts knowledge via constructing only semantically consistent subclasses of homogeneous classes of objects and then filters them according to the attribute and dependency queries, retrieving knowledge. In addition, we introduced the decomposition consistency coefficient, which allows estimation of how much the algorithm can reduce the search space for knowledge extraction and improves the performance. To demonstrate some possible application scenarios for the improved algorithm, we provided an appropriate example of knowledge extraction and retrieval via decomposition of a particular homogeneous class of objects.Prombles in programming 2022; 3-4: 139-153