Artificial Intelligence

University Politehnica of Bucharest

2008-2009

Adina Magda Florea
http://turing.cs.pub.ro/aifils_08
Lecture No. 4
Knowledge representation in AI
Symbolic Logic
Simbolic logic representation
Formal system
Propositional logic
Predicate logic
Theorem proving
1. Knowledge representation
Why Symbolic logic
Power of representation
Formal language: syntax, semantics
Conceptualization + representation in a
language
Inference rules
2. Formal systems
O formal system is a quadruple
A rule of inference of arity n is an association:

Immediate consequence
Be the set of premises


An element
is an immediate consequence of a set of premises I
Re9
R , y = y ,..., y x, x, y i =1, n
n
1 n
R
i
_ ( ) e F F F,
S =< A, , , > F A 9
I = {y , ... , y
1 n
} E =
0
I A
E = E x| y E , y x}
1 0 0
n
n 1
{ - e 9
>
U
E = E x| y E , y x}
2 1 1
n
n 1
{ - e 9
>
U
E (i 0)
i
>
Formal systems - cont
If then the elements of E
i
are called
theorems

Be a theorem; it can be obtained by successive
applications of i.r on the formulas in E
i
Sequence of rules - demonstration . |
S
x |
R
x

If then can be deduced from I
I |
S
x
E = ( = )
0
A I |
x E
i
e
E =
0
I A x E
i
e
3. Propositional logic
Formal language
3.1 Syntax
Alphabet
A well-formed formula (wff) in propositional logic is:
(1) An atom is a wff
(2) If P is a wff, then ~P is a wff.
(3) If P and Q are wffs then P.Q, PvQ, PQ si PQ are wffs.
(4) The set of all wffs can be generated by repeatedly applying rules
(1)..(3).
3.2 Semantics
Interpretation
Evaluation function of a formula
Properties of wffs
Valid / tautulogy
Satisfiable
Contradiction
Equivalent formulas
Semantics - cont
A formula F is a logical consequence of a formula
P
A formula F is a logical consequence of a set of
formulas P
1
,P
n

Notation of logical consequence P
1
,P
n
F.
Theorem. Formula F is a logical consequence of a
set of formulas P
1
,P
n
if the formula P
1
,P
n
F
is valid.
Teorema. Formula F is a logical consequence of a
set of formulas P
1
,P
n
if the formula P
1
. . P
n
. ~F is a contradiction.
Equivalence rules
Idempotenta P P P P P P
Asociativitate (P Q) R P (Q R) (P Q) R P (Q R)
Comutativitate P Q Q P P Q Q P P Q Q P
Distributivitate P (Q R) (P Q) (P R) P (Q R) (P Q) (P R)
De Morgan ~ (P Q) ~ P ~ Q ~ (P Q) ~ P ~ Q
Eliminarea
implicatiei
P Q ~ P Q
Eliminarea
implicatiei duble
P Q (P Q) (Q P)
v .
v v v v . . . .
v v . .
. v . v . v . v . v
v . . v
v
.
3.3 Obtaining new knowledge
Conceptualization
Reprezentation in a formal language
Model theory
KB || x M
Proof theory
KB |
S
x M
Monotonic logics
Non-monotonic logics
3.4 Inference rules
Modus Ponens
Substitution
Chain rule

AND introduction

Transposition
P Q
Q R
P R

P
Q
P Q .
P Q
~ Q ~ P

P
P Q
Q

Example
Mihai has money
The car is white
The car is nice
If the car is white or the car is nice and Mihai
has money then Mihai goes to the mountain
B
A
F
(A v F) . B C
4. First order predicate logic
4.1 Syntax
Be D a domain of values. A term is defined as:
(1) A constant is a term with a fixed value
belonging to D.
(2) A variable is a term which may take values in
D.
(3) If f is a function of n arguments and t
1
,..t
n
are
terms then f(t
1
,..t
n
) is a term.
(4) All terms are generated by the application of
rules (1)(3).
Predicates of arity n
Atom or atomic formula.
Literal
A well formed formula (wff) in first order predicate logic is
defined as:
(1) A atom is an wff
(2) If P[x] is a wff then ~P[x] is an wff.
(3) If P[x] and Q [x] are wffs then P[x].Q[x],
P[x] vQ[x], PQ and PQ are wffs.
(4) If P[x] is an wff then x P[x], -x P[x] are wffs.
(5) The set of all wffs can be generated by repeatedly
applying rules (1)..(4).

Syntax PL - cont
Syntax - schematically
Constante Variabile Functii
a x f(x, a)
Termeni Predicate
P
Formule atomice
P(a, x)
Formule atomice negate
~P(a, x)
Literali Cuantificatori Conectori logici
-,
~, ., v,,
Formule bine formate
CNF, DNF
Conjunctive normal form (CNF)
F
1
. .F
n
,
F
i
, i=1,n
(L
i1
v vL
im
).
Disjunctive normal form (DNF)
F
1
v vF
n
,
F
i
, i=1,n
(L
i1
. .L
im
)
The interpretation of a formula F in first order
predicate logic consists of fixing a domain of
values (non empty) D and of an association of
values for every constant, function and predicate
in the formula F as follows:
(1) Every constant has an associated value in D.
(2) Every function f, of arity n, is defined by the
correspondence where

(3) Every predicate of arity n, is defined by the
correspondence
D D
n

D = (x ,..., x )|x D,..., x D}

n
1 n 1 n
{ e e
} {t, D : P
n
f
4.2 Semantics of PL
a
2
f (1) f (2)
2 1
A(2,1) A(2, 2) B(1) B(2) C C D D ( ) ( ) ( ) ( ) 1 2 1 2
a f a f a f f a
(( ) ) a f a f v .
(( ) ) f a f a v .
X=1

X=2
( x)(((A(a, x) B(f (x))) C(x)) D(x)) v .
D={1,2}
Interpretation - example
4.3 Properties of wffs in PL
Valid / tautulogy
Satisfiable
Contradiction
Equivalent formulas
A formula F is a logical consequence of a formula P
A formula F is a logical consequence of a set of formulas
P
1
,P
n

Notation of logical consequence P
1
,P
n
F.
Theorem. Formula F is a logical consequence of a set of
formulas P
1
,P
n
if the formula P
1
. . P
n
F is valid.
Teorema. Formula F is a logical consequence of a set of
formulas P
1
,P
n
if the formula P
1
. . P
n
. ~F is a
contradiction.
Echivalenta cuantificatorilor
(Qx)F[x] G (Qx)(F[x] G) (Qx)F[x] G (Qx)(F[x] G)
~ (( x)F[x]) ( x)(~ F[x]) ~ (( x)F[x]) ( x)(~ F[x])
( x)F[x] ( x)H[x] ( x)(F[x] H[x]) ( x)F[x] ( x)H[x] ( x)(F[x] H[x])
(Q x)F[x] (Q x)H[x
1 2
v v . .
- -
. . - v - - v
. ] (Q x)(Q z)(F[x] H[z]) (Q x)F[x] (Q x)H[x] (Q x)(Q z)(F[x] H[z])
1 2 1 2 1 2
. v v
Equivalence of quantifiers
Examples
All apples are red
All objects are red apples
There is a red apple
All packages in room 27 are smaller than any package in
room 28
All purple mushrooms are poisonous
x (Purple(x) . Mushroom(x)) Poisonous(x)
x Purple(x) (Mushroom(x) Poisonous(x))
x Mushroom (x) (Purple (x) Poisonous(x))
(x)(-y) loves(x,y)
(-y)(x)loves(x,y)
4.4. Inference rules in PL
Modus Ponens

Substitution
Chaining
Transpozition
AND elimination (AE)
AND introduction (AI)
Universal instantiation (UI)
Existential instantiation (EI)
Rezolution

Example
Horses are faster than dogs and there is a greyhound that is faster than
every rabbit. We know that Harry is a horse and that Ralph is a rabbit.
Derive that Harry is faster than Ralph.
Horse(x) Greyhound(y)
Dog(y) Rabbit(z)
Faster(y,z))





y Greyhound(y) Dog(y)
x y z Faster(x,y) . Faster(y,z) Faster(x,z)
x y Horse(x) . Dog(y) Faster(x,y)
-y Greyhound(y) . (z Rabbit(z) Faster(y,z))
Horse(Harry)
Rabbit(Ralph)
Proof example
Theorem: Faster(Harry, Ralph) ?
Proof using inference rules
1. x y Horse(x) . Dog(y) Faster(x,y)
2. -y Greyhound(y) . (z Rabbit(z) Faster(y,z))
3. y Greyhound(y) Dog(y)
4. xyz Faster(x,y) . Faster(y,z) Faster(x,z)
5. Horse(Harry)
6. Rabbit(Ralph)
7. Greyhound(Greg) . (z Rabbit(z) Faster(Greg,z)) 2, EI
8. Greyhound(Greg) 7, AE
9. z Rabbit(z) Faster(Greg,z)) 7, AE
10. Rabbit(Ralph) Faster(Greg,Ralph) 9, UI
11. Faster(Greg,Ralph) 6,10, MP
12. Greyhound(Greg) Dog(Greg) 3, UI
13. Dog(Greg) 12, 8, MP
14. Horse(Harry) . Dog(Greg) Faster(Harry, Greg) 1, UI
15. Horse(Harry) . Dog(Greg) 5, 13, AI
16. Faster(Harry, Greg) 14, 15, MP
17. Faster(Harry, Greg) . Faster(Greg, Ralph) Faster(Harry,Ralph)
4, UI
18. Faster(Harry, Greg) . Faster(Greg, Ralph) 16, 11, AI
19. Faster(Harry,Ralph) 17, 19, MP
Proof example - cont