Using Markov Chain Monte Carlo to play Trivia

《Using Markov Chain Monte Carlo to play Trivia》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、UsingMarkovChainMonteCarlotoPlayTriviaDanielDeutch1OhadGreenshpan1,2BorisKostenko1TovaMilo11Tel-AvivUniversity2IBMHaifaResearchLabsAbstractWeintroduceinthisDemonstrationasystemcalledsupplydatainvariousareas.Thegameisruninrounds,inTriviaMassterthatgeneratesaverylargeDatabaseoffactsinaeachofwhic

2、husersaregivenaquestion(e.g.namesofworldvarietyoftopics,andusesitforquestionanswering.Thefactsarecapitalcities;or“WhatisthecapitalcityofChina?”)andarecollectedfromhumanusers(thecrowd);thesystemmotivatesrequiredtoprovideoneormoreanswers.TheanswersgivenuserstocontributetotheDatabasebyusingaTrivi

3、aGame,whereusersgainpointsbasedontheircontribution.AkeychallengebytheusersarestoredintheDatabase.TofurthermotivatehereistoprovideasuitableDataCleaningmechanismthatuserstocontributecorrectandnewpiecesofinformation,weallowstoidentifywhichofthefacts(answerstoTriviaquestions)useascoringmechanism,a

4、ssigninghighscorestousersthatsubmittedbyusersareindeedcorrect/reliable,andconsequentlycontributeusefulfacts,aswellasamechanismthatdecideshowmanypointstograntusers,howtoanswerquestionsbasedwhichquestionstoask.ThesecondcomponentisaQueryonthecollecteddata,andwhichquestionstopresenttotheTriviaplay

5、ers,inordertoimprovethedataquality.AsnoexistingAnsweringmechanism.Itreceivesasinputa(form-based)singleDataCleaningtechniqueprovidesasatisfactorysolutionqueryoverthecollectedData,anddecideswhatisthecorrecttothischallenge,weproposehereanovelapproach,basedonanswer,basedondatacleaningrules.adeclar

6、ativeframeworkfordefiningrecursiveandprobabilisticManyapproachesforDataCleaninghavebeenproposedinDataCleaningrules.Oursolutionemploysanalgorithmthatistheliterature.Forexample,asimpleapproachdecidesbetweenbasedonMarkovChainMonteCarloAlgorithms.I.INTRODUCTIONtwocontradictingfactsaccordingtotheirs

7、upport[4];anotherapproachsuggeststheapplicationof“transformation”rulesHarnessingacrowdofusersforthecollectionofmassdata[5]thatfixpartsofthedata.Arecentpaper[6]suggestsandforsolvingproblemshasrecentlybecomeawide-spreadtograduallycleandata