TITLE:
Knowledge Graph Application in KM for Accounting and Supply Chain in SMEs
AUTHORS:
Yaxin Zheng
KEYWORDS:
Knowledge Graph, Large Language Model, BERT, Knowledge Management, Small and Medium-Sized Enterprises, Accounting, Supply Chain Management
JOURNAL NAME:
Open Journal of Business and Management,
Vol.14 No.1,
January
21,
2026
ABSTRACT: In today’s data-driven business environment, small and medium-sized enterprises (SMEs) struggle to implement effective knowledge management (KM) due to limited financial, technical, and human resources. This study proposes an intelligent KM framework that integrates large language models (LLMs), that is, BERT-based models and knowledge graphs (KGs) to support accounting and supply chain decision-making in SMEs. Using a medical device manufacturer as a case enterprise, unstructured texts from accounting vouchers, ERP operation logs and supply chain records are collected, cleaned, anonymised, and manually annotated with entities and relations via Doccano. An end-to-end architecture based on BERT is adopted: for named entity recognition (NER) two models (BERT + CRF and BERT + BiLSTM + CRF) are trained, while for relation extraction (RE) two models (BERT and BERT + BiLSTM) are compared. Experimental results show that BERT + BiLSTM + CRF achieves the best NER performance (Precision 0.73, Recall 0.76, F1-score 0.74), and BERT attains superior RE performance (Precision 0.69, Recall 0.80, F1-score 0.74). The optimal models are then used to automatically construct RDF-like entity-relation-entity triples, which are stored and visualized in the Neo4j Aura cloud graph database via Py2neo. The resulting enterprise KG supports semantic querying, knowledge discovery, and interactive Q&A for accounting and supply chain tasks, and can be further leveraged for personalized training and internal audit support. The findings demonstrate that an LLM-enabled KG approach can provide a cost-effective, scalable KM solution for SMEs, while also highlighting challenges related to data quality, computational resources, and ongoing KG maintenance.