Relationship Extraction and Processing for Knowledge Graph of Welding Manufacturing

Abstract

Acquiring welding domain relationships and forming a knowledge graph can positively impact complex engineering problem solving and intelligent manufacturing applications. However, relationships are lacking in the welding domain. The relationship extraction and processing solution are designed to handle data with different characteristics in welding fabrication. The BiLSTM+Attention and CR-CNN models are employed to extract relations in unstructured documents. The neighborhood rough setbased association rule model is proposed for project-specific documents to accomplish relationship acquisition, in which invalid attributes are removed via neighborhood rough sets and attribute values are related via association rules. In addition, the knowledge graph is built based on extracted relationships, and unique empirical relationships are handled by introducing relational nodes and databases. The results show that BiLSTM+Attention gets a good score with Macro-average metrics (0.788 for Precision, 0.846 for Recall, and 0.816 for F1-score). The relational rules obtained via the proposed model are consistent with the production experience. The constructed knowledge graph effectively handles empirical relationships while positively impacting knowledge retrieval, intelligent question and answer, and decision-making for complex engineering problems.

A Review and Prospects of Manufacturing Process Knowledge Acquisition, Representation, and Application

“…Ye et al [139] developed a CNC machining knowledge base using cloud technology and web ontology language and realized cloud-based process solution acquisition through the NoSQL database and MapReduce model. Guan et al [140] explored welding knowledge graph building, using the BiLSTM + attention model and CR-CNN model to extract relationships from documents and applying association rule models to process specific relationships. Xiao et al [141] comprehensively analyzed the key techniques of process knowledge graphs and considered their integration with large-scale language models.…”

A Deep Neural Network-Based Multimodal Relation Extraction Model for Power Equipment

“…Relational extraction [4] is to obtain the information 3-tuple (head entity, relationship, tail entity) by identifying semantic relationships between entities in unstructured text, which is the basic task of information extraction and can support a large number of NLP downstream tasks, e.g., knowledge graph completion [5], question and answer system [6], dialogue generation [7] and so on. Early relationship extraction is to obtain various relationships between entities by pattern matching, but the rules are difficult to specify and cannot be migrated to other domains, and it has obvious defects.…”

“…In the model for entity relationship extraction, the input to the BLSTM network consists of a collocation of BERT-generated word vectors, image structural features, and feature vectors fused with textual semantic features. In order to make full use of its alignment information, semantic and structural similarity feature matrices are fused for deciding the most relevant visual nodes, and the visual representation based on the multimodal alignment is derived to fuse them, and the multimodal feature fusion is shown in equation (5).…”

Semantic Maps for Knowledge Graphs: A Semantic-Based Summarization Approach

“…Recently, academic and private organizations have constructed KGs, such as YAGO [3], DBPedia [4], Freebase [5], NELL [6], Google Knowledge Graph [7], Microsoft Satori [8], Facebook Entity Graph [9], and Wikidata [10], which contain millions of entities and billions of relationships. The main applications of KGs include the enhancement of search engines like Google [7] or Bing [8], question answering [11], information retrieval, recommender systems [12], [13], domain specific KG building [14]- [16], and decision support in the life sciences [17]- [20].…”