四川省数字孪生流域知识图谱的设计与实现

Design and application of knowledge graph in digital twin watersheds development in Sichuan Province

摘要:

为解决水利信息资源管理分散、利用率低、缺乏统一整合和关联应用等问题，满足数字孪生流域建设中知识平台的构建需求，以数字孪生流域建设的数据底板为基础，依托自顶向下和自底向上相结合的方法，通过总体框架设计，基于 BERT-BiLSTM-CRF 命名实体识别、Transformer 模型的关系抽取、Neo4j 的存储和可视化展示完成知识图谱的搭建，并基于知识图谱建设性构建智能问答系统，为四川省数字孪生流域物理流域的全映射提供支撑， 实现知识的可视化表达、精准查询与智能推荐，为四川省河湖智能管理提供知识性参考。

Abstract:

In order to solve the problems of segmented information，insuffi cient utilization，lack of integration and disconnected applications，as well as to meet the needs of knowledge graph platform in digital twin watersheds development，this paper proposes a knowledge graph platform based on the data backplane with information from digital-twin watersheds，in both top-down and bottom-up ways. The platform is built by general framework design with a combination of the BERT-BILSTM-CRF model for entity recognition，Transformer model for relation extraction and Neo4j model for visualizing the knowledge output. A knowledge-based question answering system is also established to visualize the knowledge，accurately match questions with information in the data backplane and generate intelligent recommendations. The platform has mapped all watersheds in Sichuan province，supporting digital twin watersheds development and contributing to smart governance of rivers and lakes.