长江流域取水许可知识图谱问答系统OA北大核心CSTPCD
Knowledge graph Q&A system of water intake permission based on pre-trained language model in Changjiang River Basin
随着水资源取水许可领域管理要求的不断提高,传统水资源取水许可信息管理系统难以满足复杂的信息检索需求,制约了水资源精细化管理水平的提升.为了打破系统间信息孤岛,提升取水许可信息检索效率,建立了长江流域取水许可知识图谱,基于大规模预训练语言模型提出了包含实体提及识别、实体链接、关系匹配等功能的知识图谱问答流水线方法,结合取水许可领域数据特点采用BM25 算法进行候选实体排序,构建了长江流域取水许可知识图谱问答系统,并基于BS架构开发了Web客户端.实验表明:该系统在测试集上达到了90.37%的准确率,可支撑长江流域取水许可领域检索需求.
With the continuous increase of management requirements in the field of water intake permission,the traditional in-formation management system of water intake permission is difficult to meet the complex information retrieval needs,which re-stricts the improvement of meticulous management in water resources.A knowledge graph of water intake permission in the Changjiang River Basin is established to break the information silo between systems and improve the efficiency of information re-trieval in water intake permission,and a knowledge graph Q&A including entity mention recognition,entity link,relational matc-hing and other functions is proposed based on a large-scale pre-trained language model.According to the characteristics of data in water intake permission domain,BM25 algorithm is used to sort candidate entities to construct a knowledge base question an-swering system in the Changjiang River Basin,and a Web client is developed based on BS framework.The experiment shows that the system achieves an accuracy rate of 90.37%on the test set,which can support the retrieval needs in the field of water intake permission in the Changjiang River Basin.
曾德晶;张军;曹卫华;管党根;许婧;黎育朋
长江水利委员会 网络与信息中心,湖北 武汉 430010||长江水利委员会智慧长江创新团队,湖北 武汉 430010||长江水利委员会流域管理数字赋能技术创新中心,湖北 武汉 430010中国地质大学(武汉) 自动化学院,湖北 武汉 430074||复杂系统先进控制与智能自动化湖北省重点实验室,湖北 武汉 430074||地球探测智能化技术教育部工程研究中心,湖北 武汉 430074
水利科学
取水许可知识图谱预训练语言模型问答系统水资源长江流域
water intake permissionknowledge graphpre-trained language modelquestion answering systemwater re-sourcesChangjiang River Basin
《人民长江》 2024 (006)
234-239 / 6
湖北省自然科学基金创新群体项目(2020CFA031)
评论