Hydrological modeling knowledge graph construction and application

Aiming at the lack of scenario cases of knowledge organization and structured expression in the water conservancy vertical domain, and the difficulty of generic knowledge extraction models in the water conservancy vertical domain to achieve the expected accuracy and other issues, this paper takes the hydrological model scheme recommendation as an example, proposes the construction framework and process of knowledge mapping, constructs a knowledge model containing model inheritance and development relationship, application watersheds, and model accuracy, etc., and forms a multi-strategy knowdedge extraction method based on the periodical literature class of unstructured data sources, as well as rules and methods for entity alignment and knowledge fusion. The knowledge model contains model inheritance and development relationship, application basin, model accuracy, etc. A multi — strategy knowledge extraction method based on unstructured data source of journal literature is formed, as well as the rules and methods of entity alignment and knowledge fusion. Knowdedge extraction and fusion of journal literature in the field of hydrological modeling are carried out to construct a hydrological model knowledge graph instance, which contains a total of 14, 298 node entities, 39, 133 attribute entities, and 36, 254 relationships, and the accuracy, recall, and Fl value of the entity recognition are all above 90%. The visual expression and knowdedge application of the established graph instance were carried out, and the results show that; the graph realizes rapid acquisition and learning, organization and management of hydrological model knowledge based on journal literature, supports model retrieval and recommendation, improves the discovery and use of hydrological model knowdedge, and has reference value for the construction of hydrological knowdedge graph in similar scenarios. © 2024 International Research and Training Center on Erosion and Sedimentation and China Water and Power Press. All rights reserved.