文档介绍：Logic Programm ing and Reasoning with
plete Information
Mic hael Gelfond
Computer Science Departmen t
Univ ersit yofT exas at El P aso
El P aso
T exas
mgelfond
cs
ep
utexas
e du
Jan uary
Abstract
The purp ose of this pap er is to expand the syn tax and seman tics
of logic programs and disjunctiv e databases to allo w for the correct
represen tation of plete information in the presence of m ultiple
extensions
The language of logic programs with classical negation
epistemic disjunction
and negation b y failure is further expanded b y
new mo dal op erators K and M
where for the set of rules T and form ula
F
K F stands for
F is kno wn to b e true b y a reasoner with a set of
premises T
and M F means
F ma y b e b eliev ed to b e true
b y the
same reasoner
Sets of rules in the extended language will b e called
epistemic sp eci
cations
W e will de
ne the seman tics of epistemic
sp eci
cations
whic h expands the seman tics of disjunctiv e databases
from
GL
and demonstrate their applicabili t y to formalization of
v arious forms monsense reasoning
In particular
w e suggest
a new formalization of the closed w orld assumption whic h seems to
b etter corresp ond to the assumption
s in tuitiv e meaning
In tro duction
As w as demonstrated in recen ty ears
traditional logic programming lan
guage pro vides a p o w erful to ol for kno wledge represen tation
Its main non
monotonic feature
negation as failure
Cla
mak es it p ossible to express
man yin teresting t yp es monsense kno wledge whic h are not readily ex
pressible in classical logic
Unlik e classical logic
ho w ev er
traditional logic
programming do es not allo w a programmer to directly represen t plete
kno wledge ab out the w orld
A consisten t classical theory partitions the set
of sen tences in to three parts
those whic h are pro v able
those whic h are
refutable
and those whic h are undecidable
A logi