资源描述:
《BOOK - 2011 - Hang Li - Learning to Rank for Information Retrieval and Natural Language Processing .pdf》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、SeriesISSN:1947-4040LISYNTHESISLECTURESONMMorganClaypoolPublishers&HUMANLANGUAGETECHNOLOGIES&CSeriesEditor:GraemeHirst,UniversityofTorontoLEARNINGTORANKFORINFORMATIONRETRIEVALANDNATURALLANGUAGEPROCESSINGLearningtoRankforInformationRetrievalandNaturalLa
2、nguageProcessingLearningtoRankforHangLi,MicrosoftLearningtorankreferstomachinelearningtechniquesfortrainingthemodelinarankingtask.LearningInformationRetrievaltorankisusefulformanyapplicationsininformationretrieval,naturallanguageprocessing,anddataminin
3、g.Intensivestudieshavebeenconductedontheproblemrecentlyandsignificantprogresshasbeenmade.Thislecturegivesanintroductiontotheareaincludingthefundamentalproblems,existingapproaches,theories,applications,andfuturework.andNaturalLanguageTheauthorbeginsbysh
4、owingthatvariousrankingproblemsininformationretrievalandnaturallanguageprocessingcanbeformalizedastwobasicrankingtasks,namelyrankingcreation(orsimplyranking)andrankingaggregation.Inrankingcreation,givenarequest,onewantstogeneratearankinglistofofferings
5、Processingbasedonthefeaturesderivedfromtherequestandtheofferings.Inrankingaggregation,givenarequest,aswellasanumberofrankinglistsofofferings,onewantstogenerateanewrankinglistoftheofferings.Rankingcreation(orranking)isthemajorprobleminlearningtorank.Iti
6、susuallyformalizedasasupervisedlearningtask.Theauthorgivesdetailedexplanationsonlearningforrankingcreationandrankingaggregation,includingtrainingandtesting,evaluation,featurecreation,andmajorapproaches.Manymethodshavebeenproposedforrankingcreation.Them
7、ethodscanbecategorizedasthepointwise,pairwise,andHangLilistwiseapproachesaccordingtothelossfunctionstheyemploy.Theycanalsobecategorizedaccordingtothetechniquestheyemploy,suchastheSVMbased,BoostingSVM,NeuralNetworkbasedapproaches.Theauthoralsointroduces
8、somepopularlearningtorankmethodsindetails.TheseincludePRank,OCSVM,RankingSVM,IRSVM,GBRank,RankNet,LambdaRank,ListNet&ListMLE,AdaRank,SVMMAP,SoftRank,BordaCount,MarkovChain,andCRanking.Theauthorexplainsseveralexampleapplicationsoflearnin