compiling defeasible inheritance networks to general logic programs

《compiling defeasible inheritance networks to general logic programs》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、CompilingDefeasible
nheritan eetworksto1Generalogi rograms2ia-HuaiYou,Xian hangWang,andiYanYuanDepartmentofComputingS ien e,UniversityofAlbertaEdmonton,Alberta,CanadaAbstra tWepresentamethodof ompilingarbitrarydefeasible(inheritan e)networkstogenerallogi

2、 programs.Weshowaone-to-one orresponden ebetweenthe redulousextensionsofadefeasiblenetworkandthestablemodelsofthetranslatedlogi program.Thisresultleadstothedis overythatanelegantqueryansweringpro edureforHorty's red-ulousextensionshadlongexisted:theabdu tivep

3、roofpro edureformulatedbyEshghiandowalskiforgenerallogi programsissoundand ompletefora y li defeasiblenetworksundertheproposedtranslation.Sin ethetranslationisfaithfultothe om-monlya eptednotionofspe i ity,itleadstoanoveltransformationalapproa h:anysemanti

4、 sdenedforgenerallogi programsyieldsanextensionsemanti sfornetworks,andanyqueryansweringpro eduredevelopedforgenerallogi programs anbeusedtoanswerqueriesfornetworksunderthesamesemanti s.Thisapproa halsoyieldsnewinsightsintothediÆ ulties onfrontingpath-basedf

5、ormalisms.Essentially,reason-ingwithlogi programsover omesthediÆ ultywithpath-basedformalismsindealingwiththeintera tionsof as adedee tsoflinksequen es,possibly ompoundedbypre-emptionofpaths.ThisdiÆ ultyhasbeenparti ularlyevidentinthepastintryingtounderstand

6、themeaningofthenetworksthatinvolve y les,bothsemanti allyandproof-theoreti ally,andinthedire tlyskepti al"approa htodefeasibleinheritan e.eywrods:Defeasibleinheritan e;ogi programming;onmonotoni reasoning.1
ntrodu tion
nheritan enetworkswereformalizedtode

7、alwithreasoninginproblemswheretaxo-nomi informationisnaturallyandreadilyavailable.Whenpla edinthe ontextof ommonsenseknowledge,su hnetworksare alleddefeasiblenetworks,sin etheirlinksmaybedefeatedwhen ontraryyetmorespe i informationispresent.1Appearedinthejou

8、rnalArti ial
ntelligen e113(1-2),pp247-268,19992Correspondingauthor.1Anearlyapproa h,whi hhasbeen alledindire tortranslational,isbasedonsomeintuitiverepresentationsofanetworkinanonmonoto