面向问答系统的答案获取方法研究与实现

《面向问答系统的答案获取方法研究与实现》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、中国科学技术大学博士学位论文面向问答系统的答案获取方法研究与实现姓名：呼大为申请学位级别：博士专业：计算机应用技术指导教师：陈恩红20080501摘要取准确答案。最后，提出一种基于负载平衡的问题推荐策略。该机制通过分析用户的信誉度以及对各个领域的兴趣度和权威性，为其推荐相关问题。同时，所有待解决的问题会依照其重要程度被依次推荐，以保证重要问题能够被优先推荐。并且，问题被尽量平摊至不同的用户，以调动每位用户的参与度。此外，紧急的问题会被推荐至积极性较高的领域专家处，以确保能够被及时回答。在上述工作基础上，本文实现了两个问答原型系统。实验结果表明所提出的方法有效提高了问答系统的效

2、率。关键词：互联网，问答系统，问题推荐，信息抽取规则，用户建模，语义模板学习AbstractABSTRACTWithther印iddeVelopmentoftheWeb，peoplecaneasilystoredata，exchangeinfo砷ationa11dshareknowle姑e．AlmostaJl灿ingcanbefoundontheWeb．Nowadays，se甜chenginehaSbecomethemostimportantsystembecauseitcanhelppeopletoextractwhattheywantfromtheIntemet．Howe

3、ver，usersarelimitedtouseseVeralkeywordstodescribetheirrequirementinsearchenginesandcanonlyobtainsomerelateddocuments．Manua¨yextractingthetaLrgetedinfo肿ation仃omtheserelateddocumentsisatimewastingjob．Hence，QuestionAnswering(QA)systemwhichfocusesonsolVingtheseproblemshaSgraduallyattractedmorea

4、ndmoreresearcherS．InQAsystems，userscanusequestionstodescribewhattheyneedandobtaintheanswerstothesequestions、vhichdonotneedtoberefined．Thisisbecause，comparedtothekeywordsusedinsearchengines，thesequestionscontainmoresemanticinfomationtodescribewhattheuserwantsmoreprecisely．QAsystemscanbecateg

5、orizedintoautomaticQAsystemsanduse卜interactiveQAsystems．TheautomaticQAsystemsusesemanticmatchingtoextracttheanswers，inwhichsemanticinformationofthequestiontargetisfirstanaIyzedandthenaIltheinformationwhichmeetstherequirementwillbeextractedastheanswers．Theuse卜interactiVeQAsystemsrelyonuserso

6、f亿ringtheanswers，inwhichthequestionwiIlberecommendedtothesuitabIeuserforanswering．Inthisthesis，wefocusontheset、VotypesofQAsystemsandresearchinthemethodsofimprovingthesemanticanalysisefncacyofquestions，promOtingt11equalityofrelateddocuments，increasingtheechospeedandrecaUofanswers，andbalancin

7、gthequestionrecommendationmechaJlism．Themainresearchareasandi肌ovationjofthisthesisareaSfollows：Firstly，weproposeasemanticpattemleamingalgorithm(SIIPU水S)inwhichaneValuationstrategynamedSemanticIdentmabilityInversePattemUniversality(SIIPU)isusedtoestimatet