Table Of ContentBUILDINLGA RGEK NOWLEDGE-BASSEYDS TEMS
RepresentaatnidoI nn ferenicnte h eC ycP roject
OTHERW ORKSB YD OUGLASL ENAT
811ilEdxip1e1Srgyt s tew1i1t1hFs r,e dreicHka yesR-oth&
DonaldA .W atermaAnd,d ison-WeslReeya,d ing,
Massachuse1t98t3s.,
K1wwlcdgeS-yBsatseie11dA11 r1 fitsci ia/l1 1tellwiigthe 11ce,
RandalDla viMsc,G raw-HiNlelw, Y ork1,9 82.
K11vwlReedpgree se1A1dtdasitoinv-1W1e,s lReeya,dn ig,
Massachuse1t9t8s8,.
The 011tEv1l1vggii1cH1aael1e 1rd'bs2vn vdek d,i tiwoint,h
distiunigshecdo lleagAudedsi,s on-WesRleea�yi,n g,
Massahcusett1s9,9 (7f orthcoming).
TECHNICPAULB LISHIANDGV ISORBYO ARD
CORPORAATNED P ROFESSIOPNUABLL ISHIGNRGO UP
SteveAnd amsonG,e neraEll ectric
MaryK ayA llenl,n tellogiIsntci.c s,
PaulA spinwalIlBM,
MarcA lanB armanH,e wletPta ckard
JeffrCelya nonD,i gitEaqlu ipmenCto rporation
JennifCeorn wayG,T EE lectrPircoadlu cts
RonF eherA,T &T Technologies
StepheKni selewiDcehl,c Eol ectronCiocrsp oration
KennetKho chbecMkc,D onnelDlo uglaCso rporation
JohnL angA,n aloDge vices
ThomasM ay,T exaIsn struments
BarbarSaa nderGse,n eraMlo torTse chniCceanlt er
PhilTiupc hinskFyor,d M otorC ompany
JohnV ittaGlT,E Laboratories
Thomas WadlPoawr,c PlaScyes tems
BUILDILNAGR GE
KNOWLEDGE-BASED
SYSTEMS
Representation and
Inference
int heC ycP roject
DOUGLAS
B.L ENAT
GUHA
R. V.
Addison-WPeusblleiys Choimnpga ny, Inc.
ReadinMga,s sachuse• tMtesn loP arkC,a lifor•n Nieaw York
DonM illOsn,t ar•i Woo kinghamE,n glan• dA msterda• mB onn
Sydne•y S ingapo•rT eo ky•o Madri•d S anJ uan
Manyo ft hed esignatiuosnesdb ym anufacturaenrds
selletrods i stingutihsehip rr oducatrse c laimeads
trademarWkhse.r et hosdee signataiopnpse airnt his
booka ndA ddison-Weswlaesay w aroef a trademark
claimt,h ed esignatihoanvseb eenp rintiendi nitial
capitlaelt te(res. gK.n,o wledgCer afotr)a lcla pitlaelt ters
(e.,gK .EE).
LibmroyfC o11grCenstsa logi11g-ii1o11-1P 1D1abilni cat
LenatD,o uglaBs.
Buildilnagr gken owledge-bsayssetde ms:
representa&t iinofne reinnct eh eC YC projeIc t
DouglaBs. L enatR,. VG.u ha.
p. cm.
Bibliograpp.h y:
Includiensd ex.
ISBN0 -201-51752-3
1.A rtifiicnitaell lige2n.cE ex.p ersty stems
(Computsecri ence3).S ystedme signI.. G uha,R . V.
IIT.i tleI.l lT.i tlCeY:C project.
Q335L.4 5 1989
006.-3 dc20 89-15091
CIP
Copyrig©h 1t990b yA ddison-WesPluebyl ishing
CompanyI,n c.
Allr ighrtess ervNeod. p arotf t hipsu blicatmiaoynb e
reproducsetdo,r eidna retriesvyaslt emor,t ransmitted,
ina nyf ormo rb ya nym eanse,l ectromneicch,a nical,
photocopyirnegc,o rdionrgo ,t herwiwsiet,h outth ep rior
writtpeenr missiooftn h ep ublishPerri.n tiendt he
UniteSdt atoefsA merica. Published simiunl taneously
Canada.
Covedre sigbny H annu sD esigAns sociates
Texdte sigbny J oycCe. W eston
SponsoriEndgi toTre:d B uswick
ProductiCoono rdinatSoarn:d rHau tchinson
Seti n1 0-poiPnatl atibnyoC ompsetI,n c.B,e verly, MA
ABCDEFGHIJ-HA-89
Firpsrti ntiDecneg,m ber 1989
CONTENTS
Acknowledgxmve nts
Prefaxcvei i
1.T heC ycP hliosop1h y
1.1T.h iInc e 1
12..O vercomBirnigt tl3e ness
13..F allBiancogkn G eneKrnaolw led5g e
14..A nalogtiozF ianrg- flung K8n owledge
14.1.. WhyA naloWgoyr kFsr equen1t0l y
14..2.H owT oA nalog1i1ze
15..T heR epresenTtraapt 1i5o n
16..O ntolovgsiK.cn aolw leEdnggei neoerrW ihnyBg u ildai ng
HugKeB DIisf fefrreonBmtu ildn iSnmgaK lBls 1 7
1..61.W hyE xpeSryts teDmosn D'its pSlyanye r1g7y
16..2.Th eP atthoF ertCiol-em ing1l9i ng
1..7 BuildCiyncg 20
17.1. . Converge2n1c e
17..2.Ta ctifcoOsrn toloEgnigcianle e2r3i ng
17...3 FLruenec Thr 1y N:a tural
Language Understanding 24
1..74.F reLeu ncThr2 y :M achiLneea rni2n6g
17..5.Tr 3y :H arWdo r(kT hIast, FNreoLe u nch2)6
2.O vervieCwy co f 28
21..T heK BI tself 28
2.2T.h eI tnerfa3c0e
2.3T.h eC ycRLe presenLtaantgiuo-angI en trodu3c5t ion
2.13.. CycILsF rame-Based 35
vii
viii Contents
2.3.O2n.T opo ft hFer amIesasP readtiecl-ccualus-like
ConsatirnLta ngua3g6e
2.3.T3h.eK inodfsF ramTehsa Etx iisnCt y c3 6
2.3.T4h.e" FielStdosr"fe odEr v eSryl ootf
EverUyni t3 9
2.3.I5n.ef renicne CyWchLa:Dt o eCsy c" Do"4?0
2.3.S6a.l iAesnpte ocftt shC eo nstrLaainngtu a4g2e
3.T heC ycRle preasteinLotanng uag4e5
3.1T.h eR atioBneahlietn hdDe e siogfCn y cL4 5
3..11. TheI nttuiioBne hiCnydc' LsD eafulRtea soning
Scheme 46
31..2.T heI nfereMnocdiunlg4e 9
3..13.T heT MSM odul5e0
31..4.Th eC ontenRteisoonl uMtoidounl5 e1
31.5.. TheC ontradDiectteicoatnni do n
ResoluMtoidounl5 e1
31..6T.h eE pistemological Level/Heuristic
LevTerlna slat5o2r
3.2D.e tialosft he FRreapmree senatnadtth iCeoy nc L
Constraint Language 54
3.12.. Detaoiftl hsFe r amLea ngua5g4e
3.12..1T.er minolaongdCy o nvent5i4o ns
3.12..2.Sl oVatl ueAsr eS etosfE ntri5e5s
3.12..3.K indosfF rames 57
3.12..4.O theFeirl dosfE acShl ot,
BesitdheVesal ue5 8
3.2.D2e.t aoiftl hsCe o nstrLaainngtu a5g9e
3.2.2S.y1n.to aftx h CeL 59
3.2.2W.h2y.H avBeo tahC onstrLaainngtu age
andaF ramLea ngua6g3e ?
3.2.2I.ne3fr.e nUcsei ntghCe L 64
3.2.P3r.e diocafAt reiGstr ye aterT woT h6a7n
3.2.3P.r1e.d iocafAt reiGstr ye aTthear2n 68
3.2.3T.h2eP. r edomniconef aL ow-Arity
RelathiiopNnsas:t uorraA lr tifa6c9t ual
3.2.3R.e3p.r eseTnertnianargny dH i gh-eArrity
AssertiinCo yncs7 0
3.2.3O.t4h.eI rs suienRs e presenRtealtaitoend
tot hCeh oiocfBe i naVrerys uHsi gher
AriPtrye dicates 75
3.3D.e atiolfsI n fereinnCc yicnLg8 2
3.13.. Inver8s3e s
Contents ix
3.3.g2e.nS !l ot8s4
3.3.R3e.f inem8e5n ts
3.3.I4n.- coVnetresxito8 n6s
3.3.T5ra.n sfersThr88o (ugh)
3.3.m6a.k esSenesnetFroyrai,ns deA n,t ryFo8r9m at
3.31..6 #.%mutuallyDis9j4o intWith
3.3.6#.%2c.o veri94n g
3.3.6#.%3p.a rtitio9n5e dinto
3.3.. I7nehritan9c5e
3.3.t8o.C ompu9t7e
3.3.C9o.e xntseioSneatls1 01
3.130..C onstroaniS nltost1 s0 1
3.131..s loatlVueSubs1u50m es
3.3.1A2u.to matiCcl aisfsicat1i0o6n
3.133..T oeTstExpre1s0s7i ons
3.134..D emon1s0 8
3.135..O theUrs eosfC onstraints:
StructaunrdSe csr ip1t90s
3.1351...S truct1u90r es
3.3.1E5v.e2n.t St-ruScttruurcetsu ring
ofS crip1t1s1
3.136..( SimipeldHi)of rCnl auRsuel e1s21
3.137..G enelrfaflf Rhueln(e osTr h eAibrs enc1e41)
3.3.P1r8o.c edAutrtaalc hm(evnitthaAse g edna)1 61
3.3.P1l9a.u sIinebfrleenb cyDe e termina1t61i ons
3.30..P2 lausIinebfrleenb cyRe e solCvoinnsgt ra1i1n7t s
3.31..C2 omputPilnagu siobfSi cleintayr1 i1o8s
3.32..C2 oncluAsbioounItsn fereinnCc yicnLg1 22
3.33..S2 kolemiizna tCiy1oc32n
3.4G.e tting, aPnudMte ttianlIgen,vef erle nc1e62
3.14.. GettainndPg u ttui.nVsgal. u e1s2 6
3.4.P2u.t ti1n92g
3.4.W3h.y " PuttiLnegv eAlrseI" n vistiotb hlee
Use(rO ptio1n3a0l )
3.4.G4e.t ti1n33g
3.4.T5h.eA gen-daC ontOrvoeltr h e
Inefren-cTeh eM etale1v43e l
3.41..5 W.hata T Iassk ?1 53
3.4.5W.h2a.Dt o eIst MTeoaE nx ecute
aT ask?1 53
3.4.5C.o3n.d itfioCorhn oso stihnNege xTtas k
ToW orOkn 173
3.4.M6e.t alGeuviedla onfCc yec ILn feren1c93i ng
3.4.. M7etalCeovnetlir noG le tti1n04g
x Con�nb
3.4.A8n.E xampolfMe e talIenvfeolr ma1t4i1o n
3.4.M9e.t alCeovnetlor fMo le m. .1 4.3
3.140..C ycSLy ntfaoxIr n voktihnAegg end1a4 3
3.1401...D etermiWnhientghT eoUr s e
thAeg end1a4 4
3.4.1I0d.e2n.t ifyTiasnkTg y pteh e
Correspotnotd hiQenu ge ry1 45
3.141..Me talevel Contr1o4l5 in Putting
3.5N.e avtesr sSucsri uefAsf D:i gress1i4o7n
3.15.. DistinguCiyscfh rionmg
Expert Sy1s4t7e ms
3.5.D2i.f ferDeinscteisn guCiysch ifFnrigorm s t-order
PrediCcaaltceu( lFuOsP TCh)e orem
Proving S1y84s tems
4. RepersentaWthioanta:n d W hy?1 49
41..W hoG etTsoB eS omeboodryT,,h eC osmic
CookCiuet t1er4 9
41...1 Examp1lI:esS teveDoauT ghlians1g 5?1
4..12.E xamp2l:Ie sP ers(otnh eo fas leplte ople)
aT hin1g5?1
41.3.. Examp3l:Ie s4 2 981,54 6,a3T hin1g5?1
41..4.E xamp4l:Ie sD iningAtAReasT thaiunr1ga25?n t
41.5.. Examp5l:e T Ihse MetsOuatT fhiienl1gd5? 5
41.6.. Examp6l:e t IhSseu bstGanoclaedT hin1g65?
41.7.. Examp7l:Ie sM eningainIt nidsi vidualObject
ora C ollect1i95o n?
4.2. AWrheCy a tegoUrsieuefls?1 06
4.12.. SummarA yP:o lifcoDyre cidWihnegnT o M ake
aC atego1r26y
4.2.C2a.t egoorfi eSsl1 o36t s
4.3A.t tributveeVrasCluauste esg o1r4i6 es
5. GAl ismepo fC ycG'lso bOanlt olo1g71y
51..T heG eneCraatle goirnCi yecs1 37
51..1A. F ewB asic (SalnIoddt esa 1s37)
51.2.. PartitioofTn hiinng1gs57
51.3.. RepresentveedrTshuisn g
InternalMach1i67n eThing
51.4.. IndividuvaleOrbsCjuoeslc lte c(tPia1or)nt1 97
51.5.. ThosBea sOircg aniSzlionAtggs a i1n97