1、山東大學(xué)碩士學(xué)位論文摘要(i面漢語(yǔ)的詞切分是漢語(yǔ)理解的基礎(chǔ)，因而長(zhǎng)期以來(lái)一直是漢語(yǔ)信。，處‘理領(lǐng)域的一個(gè)重要課題。從目前的分詞方法來(lái)看，它們的切分精度依然滿足不了應(yīng)用的要求。根據(jù)研究，切分精度的提高依賴于對(duì)切分歧義的消解程度，而消解歧義的理想方式是采用句法語(yǔ)義知識(shí)的評(píng)測(cè)標(biāo)準(zhǔn)。這是一個(gè)高復(fù)雜度的操作，傳統(tǒng)的串行的分詞方法在實(shí)現(xiàn)上存在效率和復(fù)雜性方面的困難。本課題力圖提出一種并行的分詞方法，目的是利用并行的特點(diǎn)，使分詞和句法語(yǔ)義分析能夠充分

2、地交互，從而更有效、合理地實(shí)現(xiàn)分詞和語(yǔ)法一語(yǔ)義分析，進(jìn)一步提“”詞方法的“分精度和“分效率。丫本文在目前的書面漢語(yǔ)的分詞和分析方法基礎(chǔ)上，對(duì)基于全切分的分詞方法作了進(jìn)一步的研究。首先系統(tǒng)地討論了全切分，提出了全切分的抽象算法模型，并特別指出了全切分普遍存在的重復(fù)切分問(wèn)題。進(jìn)而，對(duì)串行全切分和并行全切分的分詞算法和剪枝方法進(jìn)行了研究，給出了串行全切分分詞方法的算法模型。最后，提出了一種并行分詞方法。該方法在輸入的同時(shí)進(jìn)行切分，并在切分的同

3、時(shí)進(jìn)行歧義識(shí)別的句法一語(yǔ)義分析，從而實(shí)現(xiàn)了輸入與切分、切分與歧義分析的并行運(yùn)算，它們互為條件、共同發(fā)展，在輸入結(jié)束時(shí)即刻得到分詞結(jié)果。本文提出了該分詞方法的體系結(jié)構(gòu)模型，設(shè)計(jì)了在M工11lSM上的并行分詞算法，并建立了單機(jī)模擬實(shí)驗(yàn)系統(tǒng)PARASEG。經(jīng)過(guò)在PARASEG系統(tǒng)上的模擬實(shí)驗(yàn)，證明該方法是一種可行而且有效的并行分詞方法。、(該方、去具有、良高白勺理論、口應(yīng)用、介值。它，if用并、于計(jì)算技術(shù)實(shí)現(xiàn)了分詞，口山東大學(xué)碩士學(xué)位論文AB

4、STRACTWordsegmentationisthefirstandafundamentalstepforChineseinformationprocessingandisalwaysoneofthemostimportanttopicsofChineseProcessing.Buttheaccuracyofthecurrentmethodsofwordsegmentationcannotsatisfytherequirementof

5、theapplication.Asitispointedtheaccuracyofsegmentationdependsonthemethodofworddisambiguation.Oneofthebestmethodsofworddisambiguationisthroughthesyntacticsemanticanalysiswhichhasahighcomplexityandthepresentsimplemethodsofs

6、egmentationisdificulttorealize.Toraisetheaccuracyandefficiencyofwordsegmentationthispaperpresentsamethodofparallelwordsegmentation.Itcanmaketheprocessingofwordsegmentationandsyntacticandsemanticanalysismoreefficientbymak

7、inguseoftheirinteractionswitheachother.Basedoncurrentmethodsofwordsegmentationandanalysismethodsofomniwordsegmentationarediscussedinthispaper.FirstlytheproblemofwordOmniSegmentationisdiscussed.Thenamodelofwordomnisegment

8、ationisproposedandtheproblemofredundantsegmentationexistedgenerallyintheprocessofomnisegmentationispointedout.Basedontheabovethealgorithmofsequentialandparallelwordomnisegmentationandthemethodofearlypruningofunacceptedse

9、gmentationformsarediscussedandamodelofsequentialwordomnisegmentationalgorithmisgiven.InthiswayanovelparallelapproachforChinesewordsegmentationisproposed.Theapproachperformsthesegmentationwhiletheinputisenteringandatthesa

10、metimetheambiguityrecognitionwithsyntacticandsemanticanalysisisachieved.Inthiswaytheparallelofinputandsegmentationthesegmentationandsyntacticandsemanticanalysisisrealizedandthecorrectsegmentationisobtainedattheendofinput

11、.ThesystemarchitectureandtheparallelalgorithmfortheapproacharegiveninthispaperandasimulationsystemPARASEGisconstructed.ThesimulationofthemethodonasingleCPUhardwarehasprovedthattheideaproposedinthispaperisbothfeasibleande

12、ffective.Theapproachhashighvaluebothintheoryandapplication.Itadoptsthecurrentparallelcomputingtechnologyfortheachievementoftheparallelimplementationofwordsegmentationandsyntacticsemanticanalysis.Inthiswaytheprocessofword