| Abstract | Classification societies around the world have a wide range of regulations for the safe operation of ships, and in recent years, these have become increasingly detailed. Because the regulations are vast and complex, it is difficult to obtain the desired results through keyword-based searches alone. As a result, it takes a lot of time to analyze all the regulations. In addition, users unfamiliar with regulatory terminology may need to learn the exact keywords to use to get the information. Therefore, this study proposed a search method to enable effective searching of the regulations. The proposed method preprocesses the regulations, converts them to a searchable form, computes their semantic similarity to the user search query, and serves the regulations with the highest similarity as answers. By extracting basic elements such as strings from the regulations in the PDF file, we built a dataset that can represent features such as hierarchical relationships within documents. We selected the final model for the efficient search by comparing and analyzing natural language processing (NLP) models for the semantic similarity-based search of the dataset and query for search. The proposed method was applied to the regulations of the Korean Register (KR-Rules), and the results show that it can provide users with rules that show high semantic similarity to the query in a short time. |
|---|---|
| Publication Date | 2024-09-10 |
In-Su Han, Myung-Il Roh, Min-Chul Kong, Jeongyoul Lee, Seoyoon Park, "A Search Method for Ship Regulations Considering Document Features", Proceedings of ICCAS(International Conference on Computer Applications in Shipbuilding) 2024, Genoa, Italy, 2024.09.10-12