google搜索引擎方案

《google搜索引擎方案》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、www.semtrain.com.cnwww.semtrain.cnTheAnatomyofaLarge-ScaleHypertextualWebSearchEngineSergeyBrinandLawrencePageComputerScienceDepartment,StanfordUniversity,Stanford,CA94305,USAsergey@cs.stanford.eduandpage@cs.stanford.eduAbstractInthispaper,wepresentGoogle,aprototypeofalarg

2、e-scalesearchenginewhichmakesheavyuseofthestructurepresentinhypertext.GoogleisdesignedtocrawlandindextheWebefficientlyandproducemuchmoresatisfyingsearchresultsthanexistingsystems.Theprototypewithafulltextandhyperlinkdatabaseofatleast24millionpagesisavailableathttp://google

3、.stanford.edu/Toengineerasearchengineisachallengingtask.Searchenginesindextenstohundredsofmillionsofwebpagesinvolvingacomparablenumberofdistinctterms.Theyanswertensofmillionsofquerieseveryday.Despitetheimportanceoflarge-scalesearchenginesontheweb,verylittleacademicresearch

4、hasbeendoneonthem.Furthermore,duetorapidadvanceintechnologyandwebproliferation,creatingawebsearchenginetodayisverydifferentfromthreeyearsago.Thispaperprovidesanin-depthdescriptionofourlarge-scalewebsearchengine--thefirstsuchdetailedpublicdescriptionweknowoftodate.Apartfrom

5、theproblemsofscalingtraditionalsearchtechniquestodataofthismagnitude,therearenewtechnicalchallengesinvolvedwithusingtheadditionalinformationpresentinhypertexttoproducebettersearchresults.Thispaperaddressesthisquestionofhowtobuildapracticallarge-scalesystemwhichcanexploitth

6、eadditionalinformationpresentinhypertext.Alsowelookattheproblemofhowtoeffectivelydealwithuncontrolledhypertextcollectionswhereanyonecanpublishanythingtheywant.KeywordsWorldWideWeb,SearchEngines,InformationRetrieval,PageRank,Google地址：北京市朝阳区郎家园10号百事和大厦4层课程咨询热线：010-57222368/9

7、www.semtrain.com.cnwww.semtrain.cn1.Introduction(Note:Therearetwoversionsofthispaper--alongerfullversionandashorterprintedversion.ThefullversionisavailableonthewebandtheconferenceCD-ROM.)Thewebcreatesnewchallengesforinformationretrieval.Theamountofinformationonthewebisgrow

8、ingrapidly,aswellasthenumberofnewusersinexperiencedintheartofwebresearch.Peoplearelikelyt