I'm developing a Guess Who? game using Prolog. The mechanics of the game are very easy. A player (in this case, the human) chooses one person among many possible ones, and the other player (the computer) starts asking yes/no questions about some attributes of the person. Eventually, with the given answers, there will only be one possible person.
So far, I've been able to make a set of rules and predicates so that the computer can guess the person based on the questions that have been asked so far. I have a set of suspects - those are the people that, with the available clues, could fit.
suspect('Person 1') :- eyes(blue) , age(old) , gender(male).
The predicates for the attributes are defined so that they will be true either if the question regarding that attribute has not been asked yet, or if the question has been asked and the answer matches the attribute of the suspect.
gender(X) :- not(asked_gender) ; value_of(gender, X).
That way, if two suspects share the same eyes and age and different gender, as long as the gender remains unasked, both of them will be plausible suspects.
However, the hard part now is to automate the process of asking those questions. Basically, I'm looking forward to a solution where Prolog were able to get the possible values for the attributes from the suspects' predicates, instead of listing theme somewhere else. I'm pretty sure there must be a way of doing this, given prolog is able to use the program's code as data itself.
How could I do that?
This works in SWI-Prolog :
:- dynamic value/1.
suspect('Person 1') :- eyes(blue) , age(old) , gender(male).
suspect('Person 2') :- eyes(green) , age(young) , gender(male).
suspect('Person 3') :- eyes(brown) , age(young) , gender(male).
fetch(Criterion, Value) :-
retractall(value(_)),
assert(value([])),
forall(clause(suspect(_), Body),
check(Body, Criterion)),
retract(value(Value)).
check((F, T), Criterion) :-
F =..[Criterion, V1],
retract(value(V2)),
( member(V1, V2) -> V3 = V2; V3 = [V1 | V2]),
assert(value(V3)).
check(T, Criterion).
check((_F, T), Criterion) :-
check(T, Criterion).
check((F), Criterion) :-
F =..[Criterion, V1],
retract(value(V2)),
( member(V1, V2) -> V3 = V2; V3 = [V1 | V2]),
assert(value(V3)).
check((_F), _Criterion).
For example :
?- fetch(gender, Value).
Value = [male].
?- fetch(eyes, Value).
Value = [brown,green,blue].
Well, I would imagine a construction like this:
go :-
findall(People,suspect(People),SuspectList),
length(SuspectList,1),
member(Perb,SuspectList),
write('It is '),write(Perb),write('!!'),nl,!.
go :-
askQuestion,
go.
Where in askQuestion/0 you'd ask questions with the read/1 predicate and assert/1 the answers.
This is where you could try and make it 'intelligent' or you could just iterate over the different questions.
Related
This is my first-day learning prolog and I would like to write a program that decides what shoes I should wear based on the weather and what kind of appointment I have at the office. The main "function" would be declared for example:
"go :- outfit(snow, 15, casual, F1), write(F1)."
Where snow is the weather, 15 is the temperature(not relevant now), and casual is the formality of the appointment. "write(F1)" will display the "output" so the variable F1 needs to be the result of said relation(s). Here are the rules for what shoes to wear:
%Rules for weather
rain(Weather) :- (Weather = 'rain').
snow(Weather) :- (Weather = 'snow').
nice(Weather) :- (Weather = 'nice').
clear(Weather) :- (Weather = 'clear').
%Rules for formality
formal(Appointment) :- (Appointment = 'formal').
semiformal(Appointment) :- (Appointment = 'semiformal').
casual(Appointment) :- (Appointment = 'casual').
%Rules for when to wear a type of footwear
dressShoes(Appointment) :- formal(Appointment).
boots(Appointment, Weather) :- not(formal(Appointment)), (rain(Weather);snow(Weather)).
sneakers(Appointment, Weather) :- not(formal(Appointment)), (nice(Weather);clear(Weather)).
This is where my issue is, I am not sure how to tie the last three relations to a single relation that fills the variable "F1" for my final "outfit" function. I am a C++ guy, so I would like to essentially place a string into F1 like "[sneakers]" or "[boots]" but this is part of my growing pains with prolog. Any help is much appreciated.
Some misunderstandings about Prolog I guess. This kind of rule:
rule(Variable) :- (Variable = 'value').
You don't need to quote 'value', it is already an atom. Whatever book you are reading, look up atoms. It becomes:
rule(Variable) :- (Variable = value).
You don't need the extra parentheses in the rule definition. It becomes:
rule(Variable) :- Variable = value.
You don't need the explicit unification in the body. There is nothing else happening between the head and the body. So you don't need a variable, either. It becomes:
rule(value).
Applying this to your program, I get:
rain(rain).
snow(snow).
nice(nice).
clear(clear).
%Rules for formality
formal(formal).
semiformal(semiformal).
casual(casual).
Those rules say pretty much nothing ;-)
Your example at the very top:
go :- outfit(snow, 15, casual, F1), write(F1).
Does exactly the same as just calling outfit(snow, 15, casual, F1). So what is the purpose of the "go" and the "write"? Skip them I guess.
The logic of your program: can you explain it without code? Your code is so unusual that I have to guess.
If you want to say, "If the appointment is formal, put on dress shoes", you could write it like this:
occasion_shoes(Occasion, Shoes) :-
formality_occasion(Formality, Occasion),
formality_shoes(Formality, Shoes).
formality_occasion(formal, evening_party).
formality_occasion(semi_formal, office).
formality_shoes(formal, dress_shoes).
Do you see what is going on? You match the occasion to the shoes. To do that, you look up the formality of the occasion in the table formality_occasion/2 and then match it to the formality of the shoes in the formality_shoes/2 table.
If you are struggling to model your problem, you can also read up on relational database design.
I have been working with Prolog since today, and wanted to create a simple test case. The basic idea was to have multiple sports defined, and it looks as follows:
soccer :- category(ball_sport),
check(has_11_players_per_team),
check(large_ball),
check(use_feet).
tennis :- category(ball_sport),
...
category(ball_sport) :-
check(has_a_ball).
Now I wanted to create a testcase, to see if both sports are of the ball_sport category, but have no idea to check these sports against eachother .. I thought it would be something like the code below, but it's obvious not. Is there an easy way to check these predicate categories? Thanks
both_ballsports(sport_one, sport_two) :-
has_category(sport_one, ball_sport),
has_category_sport_two, ball_sport).
It seems that first of all, you want to declaratively state attributes of a sport.
For example:
sport_attributes(soccer, [ball_sport,players(22),ball(large),use(feet)]).
sport_attributes(tennis, [ball_sport,players(2),players(4),ball(small),use(racket)]).
Note that I am relating sports to attributes. For comparison, the goals of the form check(X) you use above all seem to lack a critical argument, namely the actual sport for which they hold (or not). For example, the goal check(use_feet) either holds or not, but there is no means to qualify a unary predicate of this kind and state different facts for different sports.
Note the naming convention: We describe what each argument means, separated by underscores.
With this representation, both_ballsports/2 could look like this:
both_ballsports(Sport1, Sport2) :-
ballsport(Sport1),
ballsport(Sport2).
ballsport(Sport) :-
sport_attributes(Sport, As),
member(ball(_), As).
Sample query and answer:
?- both_ballsports(Sport1, Sport2).
Sport1 = Sport2, Sport2 = soccer ;
Sport1 = soccer,
Sport2 = tennis ;
Sport1 = tennis,
Sport2 = soccer ;
Sport1 = Sport2, Sport2 = tennis ;
false.
This can be used in all directions!
Trying to create a predicate (timePeriod/2) that calculates the time period between two dates for a specific fact. I've managed to do this by myself, but face issues when 'other answers' exist in the same list (i.e. easier to explain with examples).
I have the following knowledge-base facts;
popStar('Jackson',1987,1991).
popStar('Jackson',1992,1996).
popStar('Michaels',1996,2000).
popStar('Newcastle',2000,2007).
popStar('Bowie',2008,2010).
And the following function, calculates the time between dates for a specific fact (as per below).
Predicate (timePeriod/2) -
timePeriod(PS,X) :-
bagof((Name,Start,End),popStar(Name,Start,End),PSs),X is End-Start+1)
Using Bowie as an example; it returns X=3 (which is correct).
However, when there is repetition in the list, with more than one answer available, the predicate just states 'false'. Using the facts 'Jackson' as an example, I want to be able to calculate both of the time periods for both facts; at the same time.
So, if the predicate would work for both of the Jackson facts, the predicate timePeriod would state X=10.
Would really appreciate if anyone could suggest what to change in order for this to work correctly.
Thanks.
You probably don't quite understand what foreach/3 does. I don't think I fully understand foreach/3 either. I know for sure that it is not the same as say:
for each x in xs:
do foo(x)
Another thing: "tuples" in Prolog are not what you might expect, coming from a language like Python or Haskell. This: (a,b,c) is actually this: ','(a,','(b,c)). Much better is to use a flat term, the generic form would be triple(a,b,c). For a pair, the idiom is First-Second.
So, you can simplify your call to bagof/3 to this:
?- bagof(From-To, pop_star(Name, Start, End), Ts).
Name = 'Bowie',
Ts = [2008-2010] ;
Name = 'Jackson',
Ts = [1987-1991, 1992-1996] ;
Name = 'Michaels',
Ts = [1996-2000] ;
Name = 'Newcastle',
Ts = [2000-2007].
Once you have a list as above, you need to sum the differences, which would be maybe something like:
periods_total(Ps, T) :-
maplist(period_length, Ps, Ls),
sum_list(Ls, T).
period_length(From-To, Length) :-
Length is To - From + 1.
And then you can query like this:
?- bagof(From-To, pop_star('Jackson', From, To), Ps), periods_total(Ps, T).
Ps = [1987-1991, 1992-1996],
T = 10.
?- bagof(From-To, pop_star(Name, From, To), Ps), periods_total(Ps, T).
Name = 'Bowie',
Ps = [2008-2010],
T = 3 ;
Name = 'Jackson',
Ps = [1987-1991, 1992-1996],
T = 10 ;
Name = 'Michaels',
Ps = [1996-2000],
T = 5 ;
Name = 'Newcastle',
Ps = [2000-2007],
T = 8.
SWI-Prolog has a nice library to handle aggregation: it builds upon standard 'all solutions' predicates like findall/3,setof/3,bagof/3, so you should first grasp the basic of these (as Boris explained in his answer). With the library, a single query solves your problem:
timePeriod(PS,X) :-
aggregate(sum(P), B^E^(popStar(PS,B,E),P is E-B+1), X).
Trying to create a predicate (timePeriod/2) that calculates the time period between two dates for a specific fact. I've managed to do this by myself, but face issues when 'other answers' exist in the same list (i.e. easier to explain with examples).
I have the following knowledge-base facts;
popStar('Jackson',1987,1991).
popStar('Jackson',1992,1996).
popStar('Michaels',1996,2000).
popStar('Newcastle',2000,2007).
popStar('Bowie',2008,2010).
And the following function, calculates the time between dates for a specific fact (as per below).
Predicate (timePeriod/2) -
timePeriod(PS,X) :-
bagof((Name,Start,End),popStar(Name,Start,End),PSs),X is End-Start+1)
Using Bowie as an example; it returns X=3 (which is correct).
However, when there is repetition in the list, with more than one answer available, the predicate just states 'false'. Using the facts 'Jackson' as an example, I want to be able to calculate both of the time periods for both facts; at the same time.
So, if the predicate would work for both of the Jackson facts, the predicate timePeriod would state X=10.
Would really appreciate if anyone could suggest what to change in order for this to work correctly.
Thanks.
You probably don't quite understand what foreach/3 does. I don't think I fully understand foreach/3 either. I know for sure that it is not the same as say:
for each x in xs:
do foo(x)
Another thing: "tuples" in Prolog are not what you might expect, coming from a language like Python or Haskell. This: (a,b,c) is actually this: ','(a,','(b,c)). Much better is to use a flat term, the generic form would be triple(a,b,c). For a pair, the idiom is First-Second.
So, you can simplify your call to bagof/3 to this:
?- bagof(From-To, pop_star(Name, Start, End), Ts).
Name = 'Bowie',
Ts = [2008-2010] ;
Name = 'Jackson',
Ts = [1987-1991, 1992-1996] ;
Name = 'Michaels',
Ts = [1996-2000] ;
Name = 'Newcastle',
Ts = [2000-2007].
Once you have a list as above, you need to sum the differences, which would be maybe something like:
periods_total(Ps, T) :-
maplist(period_length, Ps, Ls),
sum_list(Ls, T).
period_length(From-To, Length) :-
Length is To - From + 1.
And then you can query like this:
?- bagof(From-To, pop_star('Jackson', From, To), Ps), periods_total(Ps, T).
Ps = [1987-1991, 1992-1996],
T = 10.
?- bagof(From-To, pop_star(Name, From, To), Ps), periods_total(Ps, T).
Name = 'Bowie',
Ps = [2008-2010],
T = 3 ;
Name = 'Jackson',
Ps = [1987-1991, 1992-1996],
T = 10 ;
Name = 'Michaels',
Ps = [1996-2000],
T = 5 ;
Name = 'Newcastle',
Ps = [2000-2007],
T = 8.
SWI-Prolog has a nice library to handle aggregation: it builds upon standard 'all solutions' predicates like findall/3,setof/3,bagof/3, so you should first grasp the basic of these (as Boris explained in his answer). With the library, a single query solves your problem:
timePeriod(PS,X) :-
aggregate(sum(P), B^E^(popStar(PS,B,E),P is E-B+1), X).
I would need help about Prolog.
I posted my code, the problem is that i do not obtain the expected result.
I want planning actions for moving on table all blocks until is possible. To do this I prompt :
?- do(while(some(x, block(x) & -onTable(x)),pi(x,putOnTable(x))),s0,S).
I expect to see a response like :
S = do(putOnTable(e), do(putOnTable(b), do(putOnTable(c), s0)))
but Prolog returns "false" only. Someone can help me??
% Golog interpreter
%:- [golog_swi].
:- discontiguous clear/2, on/3, onTable/2.
:- op(800,xfy,[&]).
do(E,S,do(E,S)):- primitive_action(E),poss(a,S).
% Primitive Action Declarations.
primitive_action(putOn(_,_)).
primitive_action(putOnTable(_)).
poss(putOn(X,Y),S) :- clear(X,S), clear(Y,S), \+ on(X,Y,S), \+ X=Y.
poss(putOnTable(X),S):- clear(X,S), \+(onTable(X,S)).
% Successor State Axioms.
on(X,Y,do(A,S)):- A = putOn(X,Y); on(X,Y,S), \+ (A = putOnTable(X); A = putOn(X,_)).
onTable(X,do(A,S)) :- A = putOnTable(X); onTable(X,S), \+ A= putOn(X,_).
clear(X,do(A,S)) :- on(Y,X,S), (A = putOn(Y,_) ; A = putOnTable(Y)); clear(X,S), \+ A = putOn(_,X).
% Restore suppressed situation arguments
restoreSitArg(onTable(X),S,onTable(X,S)).
restoreSitArg(on(X,Y),S,on(X,Y,S)).
restoreSitArg(clear(X),S,clear(X,S)).
block(X):- member(X,[a,b,c,d,e]).
% iniTial COndition
onTable(a,s0).
on(b,a,s0).
on(c,b,s0).
clear(c,s0).
onTable(d,s0).
on(e,d,s0).
clear(3,s0).
thank you!!!
Your predicate do/3 cannot succeed because the goal primitive_action/1 will fail with your query.
Currently, while/2 is not described in primitive_action/1 and it seems it is missing also from your program. So you need to extend primitive_action/1 by further facts, or add a new rule to do/3. And in addition to that you need to describe what while/2 means.
This question is actually about Golog. Your mistake is pretty mundane: you didn't copy the Golog interpreter code into your source file/directory.
Golog defines a number of high-level programming constructs, including while-loops and non-deterministic picks (pi), used here. I'm sure you don't want to reinvent Golog, so just go and get it. I'm assuming that your question is part of an assignment of sorts, and your teacher probably pointed you to the Golog interpreter. Otherwise, you can always find it on the pages of the cognitive robotics group at the Univ. of Toronto: http://www.cs.toronto.edu/cogrobo/main/systems/index.html