Knowledge graph construction based on ship collision accident reports to improve maritime traffic safety

As an important data source, marine accident investigation reports are frequently used for accident analysis. However, it is hard to extract effective information since key knowledge is normally hidden in large blocks of text. In most cases, the collection of accident-related data is done manually. In this paper, a new knowledge graph construction approach to explore ship collision accidents is proposed, aiming to show the correlation among important factors of the accidents. In this research, 241 investigation reports on ship collision accidents from 2018 to 2021 published on the official website of the China Maritime Safety Administration (CMSA) were collected and analyzed. Then, the ship collision accident ontology module is constructed. According to the ontology information in the accident reports, entities were divided into context-based metadata and content-based metadata, which were used for describing different types of data. To extract the information for the accident report with semi-structured data, an information extraction module based on ontology was proposed. In this process, natural language processing (NLP) was used to obtain text information about the ontology. On this basis, the Ship Collision Accident Knowledge Graph (SCAKG) including 910 entity nodes and 1920 relation edges was constructed and stored in the graph database Neo4j. Finally, two case retrievals were conducted using the SCAKG to show the potential utilization of the method. The results show the effectiveness of the proposed approach in terms of discovering the internal relationship of the accident and could be used to expedite the judicial process, which simplifies the process of marine accident investigation.