Combining Deep Learning with Knowledge Graph for Design Knowledge Acquisition in Conceptual Product Design

Automatic knowledge graph creation from engineering standards using the example of formulas

“…One approach is to transform the XML content into a semantically enriched knowledge representation, while previous research has already shown that knowledge graphs (KG) can enable machine-actionability (Manoharan et al, 2019;Ehring et al, 2021). Hereby, a KG is a semantic network that interconnects data by representing entities and their relations whereas the knowledge structure is referred to as ontology (Huang et al, 2023). The KG creation from engineering standards was demonstrated using the example of mathematical equations that were automatically extracted from XML standards, transferred to graph patterns, and provided via flexible interfaces (Luttmer et al, 2021).…”

“…While different research papers on extraction of valuable information from engineering standards exist, the automatic aggregation to KGs has not been investigated in detail. Within the engineering design domain, Huang et al (2023) present an approach for design knowledge acquisition in conceptual product design. Here, a KG is constructed using five layers, namely the data resources layer, the domain ontology layer, the entity extraction layer, the relation extraction layer, and the KG application layer.…”

“…Here, a KG is constructed using five layers, namely the data resources layer, the domain ontology layer, the entity extraction layer, the relation extraction layer, and the KG application layer. To extract entities and relations, a transformer-based method is used achieving an extraction accuracy of 94.2% for entity extraction and 68.5% for relation extraction (Huang et al, 2023). The final KG is stored in the graph database Neo4j.…”

“…Based on the previously discussed literature, particularly the methodologies proposed by Huang et al (2023) and Zhao et al (2020), the automatic KG creation follows characteristics steps, namely the creation of an ontology, information extraction, the construction of the KG, and its subsequent storage. These steps are supported by the results of a systematic literature review on KG development processes, which comprises six key steps, following a top-down approach: (Tamašauskaitė and Groth, 2023) 1.…”

“…The knowledge extraction is considered as main step of the automatic KG creation process as the overall quality depends on the accuracy of entity and relations extraction (Huang et al, 2023). Previous analysis has shown that the mathematical equation itself, the used variables, their definitions as well as relations between equations are of high relevance to correctly interpret and apply formulas (Luttmer et al, 2022).…”