Semi-Automated Knowledge Graph Construction for Vietnamese E-Commerce: A BERT-Based Approach

Knowledge Graph (KG) is a database that stores information in the form of nodes and edges. The edges represent semantic relationships between each node, which makes it easier to manage massive and diverse data in the e-commerce sector. KG is constructed with two primary tasks: Named Entity Recognition (NER) to extract data stored at nodes and Relation Extraction (RE) to find the edge information of the graph. This paper proposes an end-to-end system for semi-automatically constructing knowledge graph in the Vietnamese e-commerce domain. The paper constructs Vietnamese datasets for NER and RE tasks, specifically in the e-commerce domain. The system applies a variant of BERT to extract information at the nodes and edges of the graph. Finally, the extracted data is visualized using Neo4j. Experimental results show that the system achieves an F1Score of 0.75 for the NER task and 0.84 for the RE task.