Resolution in Predicate Logic

04/09/2023

Sudeshna Sarkar
Arnab Sarkar
Koustav Rudra
Conversion to Normal Form
• A formula is said to be in clause form if it is of the form:
• ∀!"∀!#… ∀!$ [C1∧C2 ∧… ∧Ck]
• Each of these clauses could be disjunction of some predicates

• All first-order logic formulas could be converted to clausal form

• ∀! {p(x)à ∃% {~∀& 𝑞 𝑥, 𝑦 à𝑝 𝑓 𝑥1 ∧ ∀& 𝑞 𝑥, 𝑦 à𝑝 𝑥 }}

Conversion to Normal Form
• Step1: Take the existential closure and eliminate redundant quantifiers. This
introduces ∃!" and eliminates ∃% , so:

• ∀! {p(x)à ∃% {~∀& 𝑞 𝑥, 𝑦 à𝑝 𝑓 𝑥1 ∧ ∀& 𝑞 𝑥, 𝑦 à𝑝 𝑥 }}

• ∃!" ∀! {p(x)à {~∀& 𝑞 𝑥, 𝑦 à𝑝 𝑓 𝑥1 ∧ ∀& 𝑞 𝑥, 𝑦 à𝑝 𝑥 }}

Conversion to Normal Form
• Step2: Rename any variable that is quantified more than once
• y has been quantified twice, so:

• ∃!" ∀! {p(x)à {~∀& 𝑞 𝑥, 𝑦 à𝑝 𝑓 𝑥1 ∧ ∀& 𝑞 𝑥, 𝑦 à𝑝 𝑥 }}

• ∃!" ∀! {p(x)à {~∀& 𝑞 𝑥, 𝑦 à𝑝 𝑓 𝑥1 ∧ ∀% 𝑞 𝑥, 𝑧 à𝑝 𝑥 }}

Conversion to Normal Form
• Step3: Eliminate implication

• ∃!" ∀! {p(x)à {~∀& 𝑞 𝑥, 𝑦 à𝑝 𝑓 𝑥1 ∧ ∀% 𝑞 𝑥, 𝑧 à𝑝 𝑥 }}

• ∃!" ∀! {~p(x) ∨ {~∀& ~𝑞 𝑥, 𝑦 ∨𝑝 𝑓 𝑥1 ∧ ∀% ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

Conversion to Normal Form
• Step4: Move ~ all the way inwards

• ∃!" ∀! {~p(x) ∨ {~∀& ~𝑞 𝑥, 𝑦 ∨𝑝 𝑓 𝑥1 ∧ ∀% ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

• ∃!" ∀! {~p(x) ∨ {∃& 𝑞 𝑥, 𝑦 ∧ ~𝑝 𝑓 𝑥1 ∧ ∀% ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

Conversion to Normal Form
• Step5: Push the quantifiers to the right

• ∃!" ∀! {~p(x) ∨ {∃& 𝑞 𝑥, 𝑦 ∧ ~𝑝 𝑓 𝑥1 ∧ ∀% ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

• ∃!" ∀! {~p(x) ∨ { ∃& 𝑞 𝑥, 𝑦 ∧ ~𝑝 𝑓 𝑥1 ∧ ∀% ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

Conversion to Normal Form
• Step6: Eliminate existential quantifier (Skolemization)

• ∃!" ∀! {~p(x) ∨ { ∃# 𝑞 𝑥, 𝑦 ∧ ~𝑝 𝑓 𝑥1 ∧ ∀$ ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}
• Pick out leftmost ∃# B(y) and replace it by B(F(xi1, xi2,…, xin)), where:
• xi1, xi2,…, xin are all the distinct free variables of ∃# B(y) that are universally quantified to the
left of ∃# B(y), and
• F is any n-ary function constant which does not occur already

• ∀!" ∀!% ∀!& ∃# B(y, 𝑥1, 𝑥2, 𝑥3)

• ∀!" ∀!% ∀!& B(F(𝑥1, 𝑥2, 𝑥3), 𝑥1, 𝑥2, 𝑥3)
• ∀!" ∀!% ∀!& ∃# [B1(y, 𝑥1) ∧ B2(𝑥1, 𝑥2)∧ B3(𝑥2, 𝑥3)]
• ∀!" ∀!% ∀!& [B1(F(𝑥1, 𝑥2, 𝑥3), 𝑥1) ∧ B2(𝑥1, 𝑥2)∧ B3(𝑥2, 𝑥3)]
Conversion to Normal Form
• Step6: Eliminate existential quantifier (Skolemization)

• ∃!" ∀! {~p(x) ∨ { ∃& 𝑞 𝑥, 𝑦 ∧ ~𝑝 𝑓 𝑥1 ∧ ∀% ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

• ∀! {~p(x) ∨ { 𝑞 𝑥, 𝑔(𝑥) ∧ ~𝑝 𝑓 𝑎 ∧ ∀% ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}
Conversion to Normal Form
• Step7: Move all universal quantifiers to the left

• ∀! {~p(x) ∨ { 𝑞 𝑥, 𝑔(𝑥) ∧ ~𝑝 𝑓 𝑎 ∧ ∀% ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

• ∀! ∀% {~p(x) ∨ { 𝑞 𝑥, 𝑔(𝑥) ∧ ~𝑝 𝑓 𝑎 ∧ ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

• Right side we have a set of predicates that are quantified from outside
Conversion to Normal Form
• Step8: Distribute ∧ over ∨

• ∀! ∀% {~p(x) ∨ { 𝑞 𝑥, 𝑔(𝑥) ∧ ~𝑝 𝑓 𝑎 ∧ ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }}

• ∀! ∀% {[~p(x) ∨ 𝑞 𝑥, 𝑔(𝑥) ∧ ~𝑝 𝑓 𝑎 ]∧ [~p(x)∨ ~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 }

• ∀! ∀% {[~p(x) ∨ 𝑞 𝑥, 𝑔(𝑥) ] ∧ ~p(x) ∨ ~𝑝 𝑓 𝑎 ∧ [~p(x)∨~𝑞 𝑥, 𝑧 ∨𝑝 𝑥 ]}

• Right side we have a set of predicates that are quantified from outside
• Use Boolean algebra to get CNF
Resolution Refutation Proof
Procedure for Resolution
• Convert the set of rules and facts into clause form (conjunction of clauses)
• Insert the negation of the goal as another clause
• Use resolution to deduce a refutation

• If the refutation is obtained then the goal can be deduced from the set of facts and
rules

• 𝜑 : F1∧F2∧…∧FnàG
• 𝜑 : ~(F1∧F2∧…∧Fn)∨G Valid

• ~𝜑: F1∧F2∧…∧Fn∧~G Unsatisfiable

Resolution
• If Unify(zj, ~qk) = 𝜃

• z1∨… zm, q1∨…,qn

• SUBST(𝜃, z1∨…zi-1∨ zi-1 … zm ∨ q1∨… qk-1 ∨ qk+1 … ∨ qn)

Example
• Akash, Amit, and Arun are students of a school
• Every student is either wicked or is a good Cricket player, or both
• No Cricket player likes rain and all wicked students like potions
• Arun dislikes whatever Akash likes and likes whatever Akash dislikes
• Arun likes rain and potions

• Is there anyone who is good in Cricket but not in potions?

Representation in Predicate Logic
• Akash, Amit, and Arun are students of a school
• C1: Student(Akash)
• C2: Student(Amit)
• C3: Student(Arun)

• Every student is either wicked or is a good Cricket player, or both

• ∀! Student(x)àWicked(x) ∨ Cricket(x)
• C4: ~Student(x) ∨ Wicked(x) ∨ Cricket(x)
Representation in Predicate Logic
• No Cricket player likes rain and all wicked ones like potions
• ∀! Cricket(x)à~Likes(x, Rain)
• ∀! Wicked(x)àLikes(x, Potions)
• C5: ∀! ~Cricket(x) ∨ ~Likes(x, Rain)
• C6: ∀! ~Wicked(x) ∨ Likes(x, Potions)
Representation in Predicate Logic
• Arun dislikes whatever Akash likes and likes whatever Akash dislikes
• ∀! Likes(Akash, x) ⇔ ~Likes(Arun, x)
• ∀! [Likes(Akash, x) à~Likes(Arun, x)] ∧ [~Likes(Arun, x) àLikes(Akash, x)]
• C7: ∀! ~Likes(Akash, x) ∨ ~Likes(Arun, x)
• C8: ∀! Likes(Akash, x) ∨ Likes(Arun, x)

• Arun likes rain and potions

• C9: Likes(Arun, Rain)
• C10: Likes(Arun, Potions)
Representation in Predicate Logic
• Is there anyone who is good in Cricket but not in potions?
• G: ∃! Cricket(x) ∧ ~Likes(x, Potions)
• ~G: ∀! ~ Cricket(x) ∨ Likes(x, Potions)
• C11: ∀! ~ Cricket(x) ∨ Likes(x, Potions)
 C10: Likes(Arun, C7: ~Likes(Akash, x) ∨ C1: Student(Akash)
Potions) ~Likes(Arun, x) C2: Student(Amit)
{x/Potions}
C3: Student(Arun)
C12: ~Likes(Akash, C11: ~ Cricket(x) ∨ C4: ~Student(x) ∨
Potions) Likes(x, Potions) Wicked(x) ∨ Cricket(x)
{x/Akash}
C5: ∀" ~Cricket(x) ∨
C6: ~Wicked(x) ∨ ~Likes(x, Rain)
C13: ~Cricket(Akash)
Likes(x, Potions)
How to make searchC6: ∀" ~Wicked(x) ∨
{x/Akash} procedure efficient?Likes(x, Potions)
C14: ~Wicked(Akash) C7: ∀" ~Likes(Akash, x)
FALSE ∨ ~Likes(Arun, x)
C8: ∀" Likes(Akash, x)
C16: Cricket(Akash) ∨ Likes(Arun, x)

C9: Likes(Arun, Rain)

C1: Student(Akash)
C15: Wicked(Akash) ∨ C10: Likes(Arun, Potions)
Cricket(Akash)
C4: ~Student(x) ∨ {x/Akash} C11: ∀" ~ Cricket(x) ∨
Wicked(x) ∨ Cricket(x) Likes(x, Potions)
Resolution Refutation Strategies
Resolution Strategies
• Unit Resolution
• Every resolution step must involve a unit clause

• Unit clause
• A clause that does not have any OR
• It just has one predicate or its negation

• Leads to a good speedup

• Incomplete in general
• There might be cases that can’t be deduced using unit resolution but can be
deduced using other resolution methods

• Complete for Horn knowledge bases

Resolution Strategies
• Input Resolution
• Every resolution step must involve an input sentence (from the query or KB)
• All can’t be derived clauses

• In Horn knowledge bases, Modus Ponens is a kind of input resolution strategy

• Incomplete in general

• Complete for Horn knowledge bases

 C10: Likes(Arun, C7: ~Likes(Akash, x) ∨ C1: Student(Akash)
Potions) ~Likes(Arun, x) C2: Student(Amit)
{x/Potions}
C3: Student(Arun)
C12: ~Likes(Akash, C11: ~ Cricket(x) ∨ C4: ~Student(x) ∨
Potions) Likes(x, Potions) Wicked(x) ∨ Cricket(x)
{x/Akash}
C5: ∀" ~Cricket(x) ∨
C6: ~Wicked(x) ∨ ~Likes(x, Rain)
C13: ~Cricket(Akash)
Likes(x, Potions)
C6: ∀" ~Wicked(x) ∨
{x/Akash} Likes(x, Potions)
C14: ~Wicked(Akash) C7: ∀" ~Likes(Akash, x)
FALSE ∨ ~Likes(Arun, x)
C8: ∀" Likes(Akash, x)
C16: Cricket(Akash) ∨ Likes(Arun, x)

C9: Likes(Arun, Rain)

C1: Student(Akash)
C15: Wicked(Akash) ∨ C10: Likes(Arun, Potions)
Cricket(Akash)
C4: ~Student(x) ∨ {x/Akash} C11: ∀" ~ Cricket(x) ∨
Wicked(x) ∨ Cricket(x) Likes(x, Potions)
Resolution Strategies
• Linear Resolution
• Slight generalization of input resolution

• Allows P and Q to be resolved together either

• if P is in the original KB, or
• if P is an ancestor of Q in the proof tree

• Linear resolution is complete

 C1: Student(Akash)
C11: ~ Cricket(x) ∨ C4: ~Student(x) ∨
Likes(x, Potions) Wicked(x) ∨ Cricket(x) C2: Student(Amit)

{x/Akash} C3: Student(Arun)

C4: ~Student(x) ∨
C12: ~Student(Akash) ∨ Wicked(x) ∨ Cricket(x)
C1: Student(Akash)
Wicked(Akash) ∨
Likes(Akash, Potions) C5: ∀" ~Cricket(x) ∨
~Likes(x, Rain)
C6: ~Wicked(x) ∨ C6: ∀" ~Wicked(x) ∨
Likes(x, Potions) Likes(x, Potions)
C13: Wicked(Akash) ∨
Likes(Akash, Potions) C7: ∀" ~Likes(Akash, x)
{x/Akash} ∨ ~Likes(Arun, x)
C14: Likes(Akash, C7: ~Likes(Akash, x) ∨ C8: ∀" Likes(Akash, x)
Potions) ~Likes(Arun, x) ∨ Likes(Arun, x)

{x/Potion} C9: Likes(Arun, Rain)

C15: ~Likes(Arun,
Potions) C10: Likes(Arun, Potions) C10: Likes(Arun, Potions)

C11: ∀" ~ Cricket(x) ∨

FALSE Likes(x, Potions)
What is Deduction?
• Deduction is also a kind of search

• We have to search within the rules, facts, and clauses

• Find them in appropriate order in which to apply them to deduce clauses and goals
Thank You