Prolog for Artificial Intelligence 1 SWI-Prolog

Prolog for Artificial Intelligence 1

SWI-Prolog • SWI-Prolog is a good, standard Prolog for Windows and Linux • It's licensed under GPL, therefore free • Downloadable from: http: //www. swiprolog. org/ 2

What is Prolog? • A logic programming language – “Prolog” stands for “Programming in Logic” (or “Programmation en Logique” in French) • Designed in 1972 by Alain Colmerauer and Robert Kowalski (much later than Lisp) • Used in AI programming, expert systems, and computational linguistics

Logic Programming • Creates logical models that describe the world in which a problem exists • Uses first-order logic (but can also be used with propositional logic) • A subset of declarative programming – Creates a set of conditions that describe a solution space – Two phases: declaration and interpretation

Syllogisms • “Prolog” is all about programming in logic. • Aristotle described syllogisms 2300 years ago • Sample syllogism: – Socrates is a man. – All men are mortal. – Therefore, Socrates is mortal. • This is logic. Can Prolog do it? 5

Forward and backward reasoning • A syllogism gives two premises, then asks, "What can we conclude? " – This is forward reasoning -- from premises to conclusions – it's inefficient when you have lots of premises • Instead, you ask Prolog specific questions – Prolog uses backward reasoning -- from (potential) conclusions to facts 6

Syllogisms in Prolog Syllogism Socrates is a man. All men are mortal. Is Socrates mortal? Prolog man(socrates). mortal(X) : - man(X). ? - mortal(socrates). 7

Facts, rules, and queries • • Fact: Socrates is a man(socrates). Rule: All men are mortal(X) : - man(X). Query: Is Socrates mortal? mortal(socrates). Queries have the same form as facts 8

Running Prolog I • Create your "database" (program) in any editor • Save it as text only, with a. pl extension • Here's the complete program: man(socrates). mortal(X) : - man(X). 9

Running Prolog II • Prolog is completely interactive. Begin by – Double-clicking on your. pl file, or – Double-clicking on the Prolog application and consulting your file at the ? - prompt: • ? - consult('C: \My Programs\adv. pl'). – On a mac, opening a terminal window and typing swipl • Then, ask your question at the prompt: – ? - mortal(socrates). • Prolog responds: – Yes 10

Prolog is a theorem prover • Prolog's "Yes" or “true” means "I can prove it" -Prolog's "No" or “false” means "I can't prove it" – ? - mortal(plato). False. • This is the closed world assumption: the Prolog program knows everything it needs to know • Prolog supplies values for variables when it can – ? - mortal(X). X = socrates 11

Syntax I: Structures • A structure consists of a name and zero or more arguments. • Omit the parentheses if there are no arguments • Example structures: – sunshine – man(socrates) – path(garden, south, sundial) 12

Syntax II: Base Clauses • A base clause is just a structure that represents a simple fact. • Example base clauses: – debug_on. – loves(john, mary). – loves(mary, bill). 13

Syntax III: Nonbase Clauses • A nonbase clause is a structure, a turnstile (meaning IF), and a list of structures. • Example nonbase clauses: – mortal(X) : - man(X). – mortal(X) : - woman(X) – happy(X) : - healthy(X), wise(X). • The comma between structures means AND 14

Syntax IV: Predicates • A predicate is a collection of clauses with the same functor (name) and arity (number of arguments). • loves(john, mary). loves(mary, bill). loves(chuck, X) : - female(X), rich(X). 15

Syntax V: Programs • A program is a collection of predicates. • Predicates can be in any order. • Clauses within a predicate are used in the order in which they occur. 16

Syntax VI: Variables and atoms • Variables begin with a capital letter: X, Socrates, _result • Atoms do not begin with a capital letter: x, socrates • Atoms containing special characters, or beginning with a capital letter, must be enclosed in single quotes: – 'C: \My Documents\examples. pl' • Variables consisting of, or beginning with, _ are called anonymous variables. They are used when we don’t care what their value is. 17

Syntax VII: Strings are atoms • In a quoted atom, a single quote must be doubled or backslashed: – 'Can''t, or won't? ' • Backslashes in file names must also be doubled: – 'C: \My Documents\examples. pl' 18

Common problems • Capitalization is extremely important! • No space is allowed between a functor and its argument list: man(socrates), not man(socrates). • Double quotes indicate a list of ASCII character values, not a string • Don’t forget the period! (But you can put it on the next line. ) 19

Backtracking • • • loves(femi, X) : - female(X), rich(X). female(ajoke). female(asake). rich(asake). Suppose we ask: loves(femi, X). – – female(X) = female(ajoke), X = ajoke. rich(ajoke) fails. female(X) = female(asake), X = asake. rich(asake) succeeds. 20

Readings • loves(femi, X) : - female(X), rich(X). • Declarative reading: Femi loves X if X is female and rich. • Approximate procedural reading: To find an X that Femi loves, first find a female X, then check that X is rich. • Declarative readings are almost always preferred. 21

Monotonic logic • Standard logic is monotonic: once you prove something is true, it is true forever • Logic isn't a good fit to reality • If the wallet is in the purse, and the purse in is the car, we can conclude that the wallet is in the car • But what if we take the purse out of the car? 22

Nonmonotonic logic • Prolog uses nonmonotonic logic • Facts and rules can be changed at any time – such facts and rules are said to be dynamic • assert(. . . ) adds a fact or rule • retract(. . . ) removes a fact or rule • assert and retract are said to be extralogical predicates 23

Limitations of backtracking • In Prolog, backtracking over something generally undoes it • Output can't be undone by backtracking • Neither can assert and retract be undone by backtracking • Perform any necessary testing before you use write, nl, assert, or retract 24

The Notion of Unification • • • Unification is when two things “become one” Unification is kind of like assignment Unification is kind of like equality in algebra Unification is mostly like pattern matching Example: – loves(ade, X) can unify with loves(ade, asake) – and in the process, X gets unified with asake 25

Unification I • Any value can be unified with itself. – weather(sunny) = weather(sunny) • A variable can be unified with another variable. –X=Y • A variable can be unified with (“instantiated to”) any Prolog value. – Topic = weather(sunny) 26

Unification • Once a variable has been unified with a value, it continues to have that value for the rest of the clause, and can be used that way • If we have – female(X) = female(jane) • then – write(X) will write jane. 27

Writing Prolog Programs • Suppose the database contains loves(femi, X) : - female(X), rich(X). female(jane). and we ask who Femi loves, ? - loves(chuck, Woman). • female(X) finds a value for X , say, abike • rich(X) then tests whether Abike is rich 28

Clauses as Cases • A predicate consists of multiple clauses, each of which represents a “case” grandson(X, Y) : - son(X, Z), son(Z, Y). grandson(X, Y) : - son(X, Z), daughter(Z, Y). abs(X, Y) : - X < 0, Y is -X. abs(X, X) : - X >= 0. 29

Ordering • Clauses are always tried in order • buy(X) : - good(X). buy(X) : - cheap(X). cheap(‘Java 2 Complete’). good(‘Thinking in Java’). • What will buy(X) choose first? 30

Ordering II • Try to handle more specific cases (those having more variables instantiated) first. dislikes(john, bill). dislikes(john, X) : - rich(X). dislikes(X, Y) : - loves(X, Z), loves(Z, Y). 31

Basic and derived clauses • You can often choose which facts you want to be "basic" and which derived son(isaac, steven). child(X, Y) : - son(X, Y). male(isaac). child(isaac, steven). son(X, Y) : - male(X), child(X, Y). 32

Arithmetic • The equals sign, =, means “unify. ” • 2+2 does not unify with 4. • To force arithmetic to be performed, use “is”: X is 2 + 2, X = 4. • Comparisons =: = == > >= < <= also force their operands to be evaluated. • + - * / mod, when evaluated, have their usual meanings. 33

Step 1: Declaration • Creating a knowledge base (KB) of sentences describing the world • Declaring facts – fun(ai). (AI is fun!) – likes(mark, bacon). (Mark likes bacon) • Defining rules – heartbroken(X) : - loves(X, Y), not(loves(Y, X)). (X is heartbroken if X loves Y and Y does not love X… *sniff*) • Sentences end with a period

A simple Knowledge Base orbits(mercury, orbits(venus, orbits(earth, orbits(mars, sun). orbits(moon, earth). orbits(phobos, mars). orbits(deimos, mars). planet(P) : - orbits(P, sun). satellite(S) : - orbits(S, P), planet(P).

Step 2: Interpretation • Deriving new sentences from the sentences in the KB by making queries • Uses a Prolog interpreter – SWI-Prolog, GNU Prolog, etc. • Knowledge bases must first be loaded/consulted using consult(‘filename. pl’).

Some simple queries ? - consult(‘solar. pl’). % Is the moon a satellite? ? - satellite(moon). % Is the sun a planet? ? - planet(sun). % Is Uranus a planet? ? - planet(uranus). % What are the planets? ? - planet(Planet). % What objects orbit Mars? ? - orbits(X, mars). % Is the moon made of cheese? ? - made_of_cheese(moon).

Another KB example parents(william, diana, charles). parents(henry, charles). diana, parents(charles, elizabeth, philip). parents(diana, frances, edward). parents(anne, elizabeth, philip). parents(andrew, elizabeth, philip). parents(edward. W, elizabeth, philip). married(diana, charles). married(elizabeth, philip). married(frances, edward). married(anne, mark). parent(C, M) : - parents(C, M, D). parent(C, D) : - parents(C, M, D). sibling(X, Y) : - parents(X, M, D), parents(Y, M, D). % What’s wrong with this? a. ORu. Direct(C, A) : - parent(C, P), sibling(P, A). a. ORu. Marr(C, A) : - a. ORu. Direct(C, X), married(X, A). a. ORu. Marr(C, A) : - a. ORu. Direct(C, X), married(A, X). a. ORu(C, A) : - a. ORu. Direct(C, A). a. ORu(C, A) : - a. ORu. Marr(C, A).

Predicates • Used to express facts or statements about objects and their relationships • Examples: likes heartbroken orbits • Have arguments – teaches(jpv, cs 171). (2 arguments) • Names/atoms are alphanumeric, can contain underscores, begin with a lowercase letter • Meanings of predicates, arguments, and order of arguments are arbitrary (but should be descriptive and consistent!)

Variables • • • Can be used to stand for any object Are instantiated during matching Examples: Planet X Scope is statement-wide Usually used in rules, but can also be used in facts – goes_to_heaven(Dog) : - dog(Dog). (All dogs go to heaven. ) – must_perish(Thing). (All things must perish. ) • Variable names are alphanumeric, can contain underscores, must begin with an uppercase letter • Use the underscore to indicate an anonymous variable – female(X) : - mother(X, _). (All mothers are female, regardless of who the child is. )

Queries • Describe problems to be solved • Consist of goals to be satisfied – Example: father(X) : - male(X), parent(X, _). • Goals are male(X) and parent (X, _). – Goals are checked against facts in the KB – Rules are satisfied if all the goals in the “if” part (in the body, separated by commas) are satisfied – Variables can take on any value • Press semi-colon to look for further matches • Uses backtracking to satisfy goals in all possible ways (“brute-forcing” it)

Prolog and FOL • Rules are usually in the form of Horn clauses – : - is reversed • Backtracking in Prolog is known in FOL terms as backward chaining – Begins with a goal (the query) – Recursively builds a set of substitutions that satisfy the goals necessary to conclude the goal • All variables in Prolog are assumed to be universally quantified (though skolemization can still be done)