Abstract
We proposed a ship regulation search method based on semantic similarities using natural language processing (NLP) technology. First, we proposed a data preprocessing algorithm to convert ship regulations in the PDF format into a dataset for search purposes. We extracted basic elements from ship regulations, removed elements that hindered search accuracy, and generated a search dataset reflecting hierarchical structural relationships. We compared various NLP models to evaluate the semantic similarity between the generated dataset and the search queries of the workers. Then, we proposed an effective search method for ship regulations. To validate the effectiveness of the proposed method, we applied it to the technical rules for classifying the Korean Register. The results verify that the proposed search method can provide ship regulations to workers with high semantic similarity as search queries in a short time.
