Fundamentals of Developing a Hybrid Ontology-Driven System for Scientific Publication Processingcientific Publications
DOI:
https://doi.org/10.71942/r0g4-xk42Keywords:
knowledge engineering, semantic web technologies, ontology, knowledge base, knowledge graph, elementary sense, natural language interface, scientific publicationsAbstract
The problem of building a hybrid ontology-driven system for processing scientific publications is considered, which combines a structured data repository, local OWL representations of individual articles, a canonical corpus knowledge graph, and a natural language interface as an intelligent component for accessing the system's information resources. The main scientific and technical problem is that most systems for processing scientific publications provide either bibliographic search or full-text access, but do not support a holistic transition from the natural language text of the article to formally verified knowledge suitable for logical inference, explanation of results, and inter-document integration. As a result, the content of a scientific publication often remains represented at the level of unstructured text rather than in the form of controlled semantic objects related to the ontology of the subject area. The natural language interface ensures the reception, interpretation, and formal transformation of user queries formulated in natural language into queries to the relevant information resources and services of the system. An approach is proposed in which the basic unit of integration between the text of the scientific article, its formalized ontological description, and the global semantic representation is the elementary sense. Unlike the traditional interpretation in the form of an isolated RDF triple, the elementary sense is interpreted as a specialized information object that connects a simple sentence, components of the subject, predicate, and object, metadata of origin, display decisions, confidence level, and validation status. A multi-level architecture of the system is justified, which covers the level of primary sources, the document level of local representation of the article, the level of elementary senses, the level of the canonical corpus knowledge graph, and the level of origin and quality control. The feasibility of using named graphs to preserve the local context of individual publications, triple stores for semantic storage, logical inference, and SPARQL access, and structured data repositories for managing the lifecycle of artifacts, versioning, processing, and processing logs is shown. The practical value of the proposed approach lies in the formation of a holistic model of integration of text processing, ontological representation, formal validation, and natural language access to knowledge in a hybrid information environment.Downloads
Posted
2026-06-19
License
Copyright (c) 2026 Arxiv Academy
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
