Language Agnostic Ontology Extension Framework

Abstract

Ontologies are powerful structures used to define the schema and organization of knowledge. They provide a framework for precisely and explicitly organizing concepts, entities, attributes, and relationships within a domain, enabling effective data management, knowledge sharing, and intelligent decisionmaking. Knowledge graphs, on the other hand, store data in a graphical form, facilitating semantically rich and interconnected data analysis, management, and exploration. Nodes represent entities, edges depict relationships between nodes, and attributes showcase node properties. Combining the strengths of ontology, which offers schema and concept-level modeling, with the data richness of knowledge graphs, one can unlock powerful reasoning and inference capabilities that closely resemble facts and truths. The symbiotic relationship between knowledge graphs and ontology, wherein ontology serves as a blueprint for representing knowledge in the graphs, creates a robust foundation for knowledge representation. Traditionally, ontologies have been created by experts or skilled teams with domain knowledge, aiming to build comprehensive ontologies supporting holistic data representation. However, these domain-specific ontologies require manual intervention for updates and remain language-specific, making it challenging to transfer them across different language formats. To address this challenge, a crucial component is an automatic generation framework for language-independent ontologies. Such a capability would provide a data-driven approach for creating necessary ontologies. The process involves taking a text corpus as input and subsequently constructing a knowledge graph through named entity recognition. This graph is then transformed into a concept-level modeling graph, where an ID-based approach is utilized to represent the concepts. The ID-based concepts can be extended to facilitate merging and updating the ontology with new relevant information. Concurrently, this approach can also be expanded to achieve language independence, enabling the conversion of the ontology to any required language. Expressing these ontologies in mathematical terms is essential to achieve language-agnosticism, ensuring completeness, relevance, and independence from specific domains, thereby making them applicable across various linguistic contexts.