基于ChatGPT的网络空间中水文文本信息抽取技术  

Hydrological Object Information Extraction Technology in Cyberspace Based on ChatGPT

作　　者：龚世诚 艾萍 熊传圣 宋艳红 GONG Shicheng;AI Ping;XIONG Chuansheng;SONG Yanhong(College of Hydrology and Water Resources,Hohai University,Nanjing 210024,Jiangsu,China;College of Computer and Information,Hohai University,Nanjing 211100,Jiangsu,China)

机构地区：[1]河海大学水文水资源学院,江苏南京210024 [2]河海大学计算机与信息学院,江苏南京211100

出　　处：《水力发电》2024年第10期28-32,共5页Water Power

基　　金：江苏省重点研发项目(BE2020729)。

摘　　要：网络空间中水文数据的体量大、种类繁多、分布广泛、价值密度较低且质量不一,对象化的数据组织方式能够有效地表达水文对象的地理空间位置、特征属性和时间属性。利用ChatGPT进行大规模水文文本的对象化信息抽取,是水文水资源领域利用网络丰富信息资源的有效方法之一,这种方法可以自动从时间、空间和要素3个维度抽取要素类型、要素属性、时间范围、时间尺度、空间范围和空间尺度等6个特征。研究显示,ChatGPT的合理应用不仅可以有效减少信息抽取的人力成本和降低技术门槛,还初步证实了生成式人工智能在水文水资源领域的有效应用。In the cyberspace,hydrological data is characterized by large volume,variety,widespread distribution,low value density and inconsistent quality,but the objectified data organization can effectively represent the geographical spatial location,characteristic attributes and temporal attributes of hydrological objects.Applying ChatGPT for objectified information extraction from large-scale hydrological texts is an effective approach in the hydrology and water resources field to fully utilize the rich network information resources.This method can automatically extract six characteristics such as element type,element attribute,temporal scope,temporal scale,spatial scope and spatial scale from three dimensions of time,space and element.Research indicates that the rational application of ChatGPT can significantly reduce the manpower costs and lower the technical barriers in information extraction,and it preliminarily verifies the effective application of generative AI in the field of hydrology and water resources.

关 键 词：信息抽取 ChatGPT 水文数据 文本信息 人工智能 

分 类 号：P337[天文地球—水文科学] TP391[水利工程—水文学及水资源]