漢語信息處理一體化全局尋優(yōu)算法研究.pdf

1、廈門大學(xué)碩士學(xué)位論文漢語信息處理一體化全局尋優(yōu)算法研究姓名：梁雄友申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：薛永生20060501AbstractAbstractParsingisoneofthefundamentalproblemsinnaturallanguageprocessingMostnaturallanguageapplicationssuchasinformationextraction，machinetransla

2、tionandspeechrecognitionwouldbenefitfromaccurateparsingresultMoreoverresearchonparsingishelpfultodiscovertheessenceofhumanthinkingsincethelanguageisthecarrierofhumanthinkingTherefore，parsingresearchisofgreattheoreticalim

3、portanceaswellasphilosophicalsignificanceInthisthesis，theperformanceoffourclassicalparsingmodelsisanalyzedisChineseparsingThecharacteristicsofthesemodelsarestudiedsoastodiscoverthereasonoftheobviousdifferencesbetweenChin

4、eseparsingandEnglishparsingAsaresult，themaintasksofcurrentChineseparingaredeterminedas：1)howtotakeintoaccounttheChinesegrammarflexibilityandthedisambiguationinformationbetweenwordlevelandsentencelevelwhenmodeling，and，2)h

5、owtoacquirethemostfavorabledisambiguationinformationfromtherelativelysmallcorpusavailableThenewChineseparsingmodelisproposedunderthecomparisonoftheresultsofaboveclassicalmodelsInordertoobtaintheglobaloptimization，thisart

6、icleproposedmanyinnovativealgorithms：1DefineinterlinkoverlappingambiguityfieldandcompoundambiguityfieldPointouttheBugonsolvinginterlinkoverlappingambiguityandcombinatorialambiguityandcompoundambiguityfieldinChinesewordse

7、gmentationofthepresentseveralkindsofmatureChineseautomaticwordsystemsfortherepresentativeexample：InstituteofComputingTechnology，ChineseLexicalAnalysisSystem(ICTCLAS)Correctthesemanticrelevancyatouttwowordsthroughcontextu

8、alinformationbasedonHownetaswellasredefinemefeeaboutthecandidateword，andamendthealgorithmofmaximumprobabilitywordsegmentation，successfullysolvethequestionsofoverlappingambiguitycombinatorialambiguity，interlinkoverlapping

9、ambiguityandcompoundambiguityfieldsinChinesewordsegmentation2InordertoobtaintheglobaloptimizationoffivedimensionalintegratedChineseinformationprocessingofwordsegmentation，newwordsrecognizing，Part—ofspeechtagging，wordsens

10、etagging，Chineseparsing，proposed”N—shortestpath”ofViterbipart—ofspeechtaggingroughmethod3ComparetwodefinitionandcomputationalmethodofthelexicalsemanticrelevancyabouttwowordsbasedonHownet，pointoutthedifferentapplicationoc

11、casionofusingthemRegardingthestageofwordsegmentationwecanusethesecondway，asforthestageofwordsenseandsyntacticstructuredisambiguationwecanusethefirstway，4PointoutatpresentitisverydifficultofusingBigram，TrigramHMMandthegra

12、mmaticalknowledgebaseofcontemporaryChinesetosolutetheproblemsofwordsenseandsyntacticstructuredisambiguationInordertosolvetheproblemsofsyntacticstructuredisambiguation，amendthePCFGalgorithmusinglinguisticenvironmentfactor

13、sbasedonHownet5Usingthestatisticsprimarily，themleforvice—methoddefinitioncomponentpartsofasentencestructuresemanticrelevancy，andcarriesonscreeningusingittothePCFGalgorithmresult，theexperimentalresultindicatedcaneffective