GraphRAG-ASCOC: A Lightweight Framework for Adaptive Synonym-aware Clustering and Ontology Completion

Recent advances in large language models (LLMs) are accelerating the use of these models in document-rich environments. However, when organizations wish to convert lengthy, complex industrial standards into digitized knowledge that can be used in smart decision-making systems, they run into a major challenge: The AI tools struggle with the large data volumes, and sometimes generate mistakes or “hallucinations”.
In this study, the authors present an advanced method called GraphRAG-ASCOC, which allows to tackle this challenge. This method improves the manner in which complex information is efficiently compacted, allowing to convert it into more accurate, reliable, and effective databases – a key foundation of expert systems and decision-support systems.

To examine the method’s practical applicability, it was tested on a large-scale US military standard (MIL-STD-6016B). The results indicate marked improvement in reducing duplications, in knowledge organization, and in generating compact ontologies suitable for rule engines, simulators, and other expert system components.

The study demonstrates how complex industrial documents and defense standards can be converted into usable, reliable information, and lays the groundwork for developing advanced expert systems in fields such as defense, industry, and manufacturing.

READ THE PAPER