Knowledge graph for maritime pollution regulations based on deep learning methods

Abstract

Pollution prevention in the shipping industry has attracted the attention of government authorities globally. A meticulous Port State Control (PSC) inspection is an effective way to prevent pollution from ships. PSC inspections involve a variety of regulations, but currently lack effective methods to utilize these regulations to support making swift and effective decisions for maritime pollution prevention management at the frontlines. The present study developed a deep learning knowledge extraction model, the Bidirectional Encoder Representation from Transformers -Multiple Convolutional Neural Network model (BERT-MCNN), to extract needed information for frontline management from the maritime pollution-prevention-related laws, regulations, rules, and conventions in China. The results revealed that the maritime pollution prevention regulation knowledge graph can be formed based on the deep learning methods, with a precision of 92.4% and 92.7% for multi-relation extraction and named entity recognition, respectively. The knowledge graph formed by the extraction results is applied based on the graph database Neo4j, which can effectively provide knowledge queries and assist PSC officers in decision-making at the inspection scene. The method proposed in the present study can be applied to analyze the regulations for pollution prevention in China and support pollution prevention inspection.