Joint Entity-Relation Extraction for Knowledge Graph Construction in Marine Ranching Equipment

• DOI: 10.20944/preprints202503.1232.v1
• Publication Date: 2025-03-20T01:14:06Z
• Authors: Du Chen; Zhiwu GAO; Sirui Li; Xuruixue Guo; Yaqi Wu; Haiyu Zhang; Delin Zhang
• Keywords: Knowledge graph; Relationship extraction

The construction of marine ranching is a crucial component China’s Blue Granary strategy, yet the fragmented knowledge system in equipment impedes intelligent management and operational efficiency. This study proposes novel graph (KG) framework tailored for equipment, integrating hybrid ontology design, joint entity-relation extraction, graph-based storage: (1) limitations existing KG are obtained through targeted questionnaires diverse users employees; (2)A domain was constructed combined top-down bottom-up approach, defining seven core concepts eight semantic relationships; (3) Semi-structured data from enterprises standards, with unstructured literature were systematically collected, cleaned via Scrapy regular expression, standardized into JSON format, forming domain-specific corpus 1,456 annotated sentences; (4)A BERT-BiGRU-CRF model developed, leveraging contextual embeddings BERT, parameter-efficient sequence modeling BiGRU, label dependency optimization using CRF. TE+SE+Ri+BMESO tagging strategy introduced to address multi-relation extraction challenges by linking theme entities secondary entities; (4)The Neo4j-based encapsulated 2,153 nodes 3,872 edges, enabling scalable visualization dynamic updates. Experimental results demonstrated superior performance over BERT-BiLSTM-CRF, achieving 86.58% precision, 77.82% recall, 81.97% F1-score(1.94% improvement, p < 0.05) . not only pioneers first structured but also offers transferable methodology vertical extraction.