自然与语言处理review

1一.1.NLP中两种派系Rational和Statistal的基本观点和方法Symbolicapproach:Encodealltherequiredinformationintocomputer(rationalism).linguisticknowledge(staticknowledge,context-dependentknowledge).worldknowledge(uniquenessofreference,typeofnoun,situationalassociativitybetweennoun)Statisticapproach:inferlanguagepropertiesfromlanguagesamples(empiricism)CollectalargecollectionoftextsrelevanttoyourdomainForeachnoun,computeitsprobabilitytotakeacertaindeterminerP(determiner|noun)=)()mindet,(nounfreqerernounfreqGivenanewnoun,selectadeterminerwiththehighestlikelihoodasestimatedonthetrainingcorpus2.给定“自然”两个字的三种编码，判断是Big5,GB2312或者UTF-8，说明理由Reason:Big5:-thefirstbyterangesfrom0xA0-0xF9,-thesecondbyterangefrom0x40-0x7e,0xA0to0xFE-ASCIIcharactersarestillrepresentedwithasinglebyte-theMSBofthefirstbyteofaBig5characterisalways1-Big5isan8-bitencodingwitha15-bitcodespaceGB2312:-containedonlyonecodepointforeachcharacter-MSB.bit-8ofeachbyte,issetto1,andthereforebecomesa8-bitcharacter.Otherwise,thebyteisinterpretedasASCII-everyChinesecharacterisrepresentedbyatwo-bytecode.theMSBofboththefirstandsecondbytesareset自然语言处理GB2312D7D4C8BBD3EFD1D4B4A6C0EDBig5A6DBB54DBB79A8A5B342B27AUTF-8E887AAE784B6E8AFADE8A880E5A484E790863.给定五个中文词语，判断属于哪种构词法（注意英文术语)Modifiednouncompound(大人，小人，热心，水手，黑板，去年)Modifiedverbcompound(寄生，飞驰，杂居，火葬，面授，单恋)2Coordinativecompound(报告，声音，奇怪，帮助，学习，购买)Antonymouscompounds(买卖，左右，高矮，大小，开关，长短)Verb-objectcompound(放心，鼓掌，动员，司机，主席，干事)Verbcomplementcompound(进来，进去，介入，改良，打破，推翻)Subject-predicatecompound(地震，心疼，民主，自决，胆小，年轻)Noun-measurecomplementcompounds(人口，羊群，书本，花朵，枪支)Modifier-noun(情人节，小说家，加油站，大学生，金黄色)Verb-objecttri-syllabiccompound(开玩笑，吹牛皮，吃豆腐)Subject-verb-object(胆结石，鬼画符，鬼打墙)Descriptive+noun(棒棒糖，乒乓球，呼啦圈)4.给出三种StructuralAmbiguities的词语的例子(Overlapping,Combinatorial,Mixed)Overlappingambiguity(交集型歧义)[网球场，美国会]Combinatorialambiguity(组合型歧义)[才能，学生会]Mixedtype(混合型歧义)[太平洋，太平，平淡]5.Writedownthreetypesoffeatureofunknownwords-abbreviation(国考-国家公务员考试)-propername/NameEntity.namesofpeople小月月3.namesofplaces延坪岛.nameoforganization上海合作组织-derivedwords:(审计人，审计员，审计局，审计处)-compounds:(光敏感，流体力学)-Numerictypecompounds:(五月三日，八点十分，第一)第二题：信息论（1）熵是什么意思？WhatistheentropyDefinedbythesecondlawofthermodynamicsAmeasureoftheenergynotavailableforworkinathermodynamicprocessAclosedsystemalwaystendstowardsachievingastatewithamaximumofentropy1.针对Limitedsubstitutability,limitedmodifiability,Limitedextentcompositional，分别给出两个QuantitativeFeatures-synonymysubstitutionandratio-featurecharacterizesthedistributionsignificanceofhowtwowordsco-occuratdifferentpositions-thenumberofpeakco-occurrence开门，斗志昂扬2.WordNet是如何识别单词的不同意思的？HowNet和TongYiCilin又是如何识别单词的不同意思的？请区分wordnet和hownet对词语的语义进行描述的方法差异WordNet:-followdifferentgrammaticalrules4-everysynsetcontainsagroupofsynonymouswordsorcollocations-differentsensesofawordareindifferentsynsets-themeaningofthesynsetsisfurtherclarifiedwithshortdefiningglossing-synsetsareconnectedtoothersynsetsviaanumberofsemanticrelationsHowNet:-theconceptdefinitioninhownetisbasedonsememes-sememesareinastructuredmarkedlanguage-hownetconstructsagraphstructureofitsknowledgebaseontheinter-concept-therepresentationinbasedonconceptsdenotedbywordsandexpressionsinbothChineseandEnglishTongYiCilin:-thishierarchicalstructurereflectsthesemanticrelationshipbetweenwords-eachminorsemanticclusterconsistsofasetofwords-wordsunderthesameminorsemanticclustersharetheconceptofthisclass3.WordNet把单词划分成synsets，那它是怎么建立synsets之间的联系的，例如Nouns和Adjectives.synsetsareconnectedtoothersynsetsviaanumberofsemanticrelations54.Homonyms，Antonyms反义，Hypernymy上位，Hyponomy下位，Holonymy整体，当然了，考试是没有给出中文的。Homonyms（同音）:oneofagroupofwordsthatsharethesamespellingandthesamepronunciationbuthavedifferentmeaningAntonyms(反义)：differentwordshavingcontradictoryorcontrarymeaningsSynonyms(同义)：differentwordshavingsimilaroridenticalmeaningsHypernymy(上位)：thesemanticrelationofbeingsuper-ordinateorbelongtoahigherrankorclassHyponomy(下位)：thesemanticrelationofbeingsub-ordinateorbelongtoalowerrankorclassHolonymy(整体)：awordthatdefinestherelationshipbetweenatermdenotingthewholeandatermdenotingMeronym(部分)：awordthatnamesapartofalargewholeMetonymy(转指)：afigureofspeechinwhichaconceptisreferredtobythenameofsomethingcloselyassociatedwiththatconceptProposition:itreferstothemeaningofastatement解释他们的意思，并给出例子；（解释的话应该是英文，例子是中文吧）4.(好像是Thomas还是谁)关于Wordsenseambiguity的两个基本假设是什么onesensepercollocation，onesenseperdiscourse66，请区分在语言建模中平滑(smoothing)和线性插值(linearinterpolation)方法在处理零概率情况时的不同7，请解释词语消歧中常用的两个假设onesensepercollocation和onesenseperdiscourseonesensepercollocation：nearbywordsprovidestrongandconsistentcluestothesenseofatargetword,conditionalonrelativedistance,orderandsyntacticrelationshiponesenseperdiscourse:thesenseofatargetwordishighlyconsistentwithinanygivendocument-truefortopicdependentwords-nottrueforverbs8,请回答关于隐马尔可夫模型（HiddenMarkovModel）的下列问题：（1）写出马尔科夫模型的三个基本元素和三个基本问题Markovassumption:)|()...|(111tkttktXsXPXXsXP三个基本元素1.Evaluationproblem2.Decodingproblem3.Learningproblem（2）请描述Viterbi算法的基本思想，同时回答该算法是针对隐马尔可夫模型的哪个基本问题(Decodingproblem)Generalidea:7Ifbestpathendinginjksqgoesthroug