简介:WordSenseDisambiguation(WSD)istodecidethesenseofanambiguouswordonparticularcontext.MostofcurrentstudiesonWSDonlyuseseveralambiguouswordsastestsamples,thusleadstosomelimitationinpracticalapplication.Inthispaper,weperformWSDstudybasedonlargescalereal-worldcorpususingtwounsupervisedlearningalgorithmsbasedon±n-improvedBayesianmodelandDependencyGrammar(DG)-improvedBayesianmodel.±n-improvedclassifiersreducethewindowsizeofcontextofambiguouswordswithclose-distancefeatureextractionmethod,anddecreasethejammingofuselessfeatures,thusobviouslyimprovetheaccuracy,reaching83.18%(inopentest).DG-improvedclassifiercanmoreeffectivelyconquerthenoiseeffectexistinginNaive-Bayesianclassifier.ExperimentalresultsshowthatthisapproachdoesbetteronChineseWSD,andtheopentestachievedanaccuracyof86.27%.
简介:案例:日本商人"嗨"的秘密在国际谈判上,有一个典型的案例.日本一家公司就进口问题与美国厂商进行谈判.美国人进行了大量的准备工作,各种资料一应俱全.谈判一开始,急于求成的美国代表口若悬河,滔滔不绝地进行讲解,翻译忙得满头大汗.日本人埋头做笔记,仔细聆听,并不时地点头,"嗨","嗨"地随声附和.美方以为日本人认同他们的阐述,非常高兴,一路讲述了下去,最后问道:"你们觉得怎样?可以签立合同吗?"谁知,日方代表却彬彬有礼地回答"我们不是太明白".美国人一下就慌了,"你们哪块不明白?"."都有些不太明白,请再从头开始详细解释一遍好吗?"美方代表几乎要疯掉,不得已,又重复一遍.这样反复几次后,美国人大败,日本人轻松结束谈判,几乎以略夺性条件获得谈判胜利.