Page 71 - 水利学报2025年第56卷第3期
P. 71

学报(中英文),2023,21(6):574 - 585.
                [26] MORADIH,AHMADIF,FEIZID.Ahybridapproachforpersiannamedentityrecognition[J].IranianJournal
                       ofScienceandTechnology,TransactionsA:Science,2017,41(1):215 - 222.
                [27] GONGC,LIZ,XIAQ,etal.HierarchicalLSTM withchar - subword - wordtree - structurerepresentationforChi
                       nesenamedentityrecognition [J].ScienceChinaInformationSciences,2020,63(10):202102.
                [28] WANGJ,XUW,FU X,etal.ASTRAL: AdversarialtrainedLSTM- CNN fornamedentityrecognition[J].
                       Knowledge - BasedSystems,2020,197:105842.
                [29] JING,YUZ.AKoreannamedentityrecognitionmethodusingBi - LSTM- CRFandmaskedself - attention[J].
                       ComputerSpeech& Language ,2021,65:101134.
                [30] DEVLINJ,CHANGM,LEEK,etal.BERT:Pre - trainingofdeepbidirectionaltransformersforlanguageun
                       derstanding [J?OL].ArXiv,2018:48550.https:??arxiv.org?abs?1810.04805.
                [31] TIAND,LIM,REN Q,etal.Intelligentquestionansweringmethodforconstructionsafetyhazardknowledge
                       basedondeepsemanticmining [J].AutomationinConstruction,2023,145:104670.
                [32] LIMAF,SOUZAV.Alargecomparisonofnormalizationmethodsontimeseries[J].BigDataResearch,2023,
                      34:100407.
                [33] 许强,崔圣华,黄维,等.面向工程地质领域的滑坡知识图谱构建方法研究[J].武汉大学学报(信息科学
                       版),2023,48(10):1601 - 1615.
                [34] 田丹,沈扬,李明超,等.混凝土坝施工文档实体知识智能挖掘方法[J].水力发电学报,2021,40(6):
                      139 - 151.



                           Constructionandapplicationofknowledgegraphforflooddefenseand
                                               rescueofwaterinfrastructure

                                                           1,2,3
                                                                        1,2,3
                                                                                          2
                                             1,2
                             ZHANGDongliang ,ZHOUWei ,MAGang               ,WANGXudong,
                                                     1,2
                                               LIUYu ,WANGXiaomao      1,2,4
                                (1.InstituteofWaterEngineeringSciences,WuhanUniversity,Wuhan 430072,China;
                      2.StateKeyLaboratoryofWaterResourcesEngineeringandManagement,WuhanUniversity,Wuhan 430072,China;
                             3.KeyLaboratoryofRockMechanicsinHydraulicStructuralEngineeringofMinistryofEducation,
                                              WuhanUniversity,Wuhan 430072,China;
                                            4.CISPDRCorporation,Wuhan 430010,China)

                  Abstract:Theknowledgeplatform isanimportantcomponentindigitaltwinofwaterconservancy.However,
                  waterconservancyknowledgeisdispersedacrossmulti - sourcetexts,whichexhibitobviousunstructuredandfrag
                  mentedcharacteristics ,andknowledgeextractionandeffectiveutilizationfacechallenges.Toaddressestheissues
                  oflowdataqualityandunderutilizationofknowledgeinthefield,thisstudyfocusesonthetextsofflooddefense
                  andemergencyrescue,andproposesanintelligentmethodforconstructingaflooddefenseandemergencyrescue
                  knowledgegraphbyimprovingtheknowledgeextractionmodelandcombiningunstructureddataandexternalsemi -
                  structureddata.Initially ,alargelanguagemodelisemployedtoextracttermfromunstructuredtextsandconstruct
                  anontologymodelbasedontermthemes,apretrainingmoduleisusedtoenhancetextrepresentationfeatures,and
                  aconvolutionalmoduleisintroducedtoimprovetheentityknowledgeextractionmodel,andanentitydataenhance
                  mentmethodisproposedtoimprovemodelaccuracy.Thenexternalencyclopediadataisextractedtoexpandthe
                  knowledgecoveragetobuildacompleteflooddefenseandrescueknowledgegraph.Experimentalresultsdemonstrate
                  thattheproposedmodelachievesan F1scoreof89.91% inentityknowledgeextraction,significantlyoutperforming
                  baselinemodels.Finally,theapplicationmethodofknowledgegraphinthefieldofflooddefenseandrescueisin
                  troduced ,whichcanformaknowledgeenginefordigitaltwinofwaterconservancyconstruction,providingknowl
                  edgesupportforfloodcontrolresearchanddecision - making.
                  Keywords:flooddefenseandrescue;knowledgegraph;knowledgeextraction;multi - sourcedata;digitaltwin

                                                                                    (责任编辑:李 娜)

                                                                                                —  3 5 3 —
   66   67   68   69   70   71   72   73   74   75   76