Research on the Construction of Maritime Legal Knowledge Graph

As the marine industry booms, the maritime legal documents are of great importance to the maneuver on the sea. However, the traditional way of consulting the text can not meet the demand of maritime operation nowadays. This paper aims to explore a way to extract and strengthen data from maritime legal texts to better support legal question answering. To mine knowledge from unstructured maritime laws and regulations, this paper proposes a method to build the maritime legal knowledge graph. To extract information from unstructured texts, BERT+BiLSTM+CRF is used for named entity recognition. DeepKE toolkit is used for relation extraction. And to strengthen the logics between entities, heterogeneous nodes are introduced to enhance the semantic associations in the maritime legal knowledge graph. The document-enhanced knowledge graph expanded in scale, so it can better support subsequent intelligent applications.