Employing natural language processing to build a knowledge graph in the realm of additive manufacturing

Additive Manufacturing (AM) provides new product development techniques to improve the manufacturing sector. However, the Design for AM (DFAM) heavily depends on domain experts, creating a substantial obstacle for beginners, necessitating a large time investment, and limiting adaptability. Thus, effectively capitalizing on the innovative product development and fabrication abilities of AM presents a significant challenge. Aiming to solve this problem, developing a Knowledge Graph (KG) is necessary to provide an intuitive reasoning technique for DFAM. This paper utilizes the advanced capabilities of Natural Language Processing (NLP) for triple recognition, i.e., the entities and relations to develop a knowledge graph (KG). Initially, multi-source textual data for the entities and relations recognition is developed. The proposed Bidirectional Encoder Representations from Transformers (BERT) model for entities and relations recognition is used for dependence and semantic feature extraction from text data. Using the BERT, the vector cosine similarities technique links entities and relations and integrates the information sequentially to develop the KG. Furthermore, the extracted entities and relations are formulated using the ontology rules to build the KG schema, and the Relational to RDF Mapping Language (R2RML) mapping provides information to support the processing of knowledge retrieval and reasoning. Thus, this provides formalized structured entities and relations visualized as nodes in the Neo4j database. An experiment and a case-based analysis were conducted to confirm the viability of the proposed methodology, specifically in the domain of Fused Deposition Modeling (FDM) centered DFAM. The proposed method demonstrates its rationality by achieving an accuracy of 97.8 and an F1-score of 97.9.