2024
Lecture Notes in Networks and Systems, Scopus
Статьи в журналах
Vol. 1209. Pp. 70-82.
Gribova V., Shalfeeva E. SMART Standards for Industry // Lecture Notes in Networks and Systems. 2024. Vol. 1209. Pp. 70-82. Springer, Cham. https://doi.org/10.1007/978-3-031-77688-5_8
An urgent task for all areas of industry is to create documents with machine-readable content (SMART-standards) that can be analyzed not only by domain specialists, but also software services for solving various professional tasks in the Industry. Such documents should be equally understandable to both experts and software systems, which is generally a contradiction. The presence of a textual, i.e. human-readable document with its subsequent translation into a machine-understandable representation is quite a difficult task due to the presence of complex semantic connections in the domain. Modern methods of natural language analysis, including Large Language Models (LLM), are not able to accurately solve this problem. Therefore, an critical task is to develop new methods and approaches to ensure human and machine comprehension while minimizing possible errors when translating from one representation to another. The purpose of this work is to describe an approach for creating and translating into a machine-understandable representation of regulatory documents of the Industry for their further use in software services and systems. The authors propose a new approach for representing normative documents as two-level graphs of knowledge. LLM models, supplemented by a specialized adapter obtained as a result of additional training, serve as the basis for ensuring the transformation of one form into another form.