:- module(owlace_dcg, [owl_ace/2]).

/** <module> OWL ACE Definite Clause Grammar

Converts a sentence in a fragment of ACE into an axiom in
a syntactic fragment of OWL 2 without data-valued properties
and some object property axioms. And vice versa.

_|
Note that in order to make OWL expressions easier to read,
we have dropped 'Object' from the names of the
class description constructors.
I.e. 'ComplementOf' is really 'ObjectComplementOf'.
|_

_|
Note that this grammar is not connected to a lexicon. To keep
things simple, we only support a small number of content words ('John',
'man', 'thing', 'like'). 
|_

@author Kaarel Kaljurand
@version 2007-03-10
*/

%% owl_ace(+OWL:term, -ACE:list) is nondet.
%% owl_ace(-OWL:term, +ACE:list) is nondet.
%
% @param OWL is an OWL axiom in the OWL functional-style syntax (Prolog notation)
% @param ACE is an ACE sentence represented as a list of atoms
%
% An example of verbalizing an OWL axiom as an ACE sentence.
%
%==
%?- owl_ace('SubClassOf'(man, 'HasValue'('InverseObjectProperty'(like), 'John')), ACE).
%
%ACE = ['Every', man, is, liked, by, 'John', '.'] ;
%
%No
%==
%
% An example of using ACE to express OWL axioms.
% Due to coordination ambiguity there can be several solutions. The solution
% that is compatible with the ACE interpretation rules (i.e. the relative clause
% modifies to the closest preceding noun) is always the first solution.
%
%==
%?- owl_ace(OWL, ['Every', 'man', 'is', 'a', 'thing', 'that', 'likes', 'a', 'thing', 'that', 'is', 'John', 'or', 'that', 'is', 'John', '.']).
%
%OWL = 'SubClassOf'(man, 'IntersectionOf'('owl:Thing', 'SomeValuesFrom'(like, 'IntersectionOf'('owl:Thing', 'UnionOf'('OneOf'('John'), 'OneOf'('John')))))) ;
%
%OWL = 'SubClassOf'(man, 'IntersectionOf'('owl:Thing', 'UnionOf'('SomeValuesFrom'(like, 'IntersectionOf'('owl:Thing', 'OneOf'('John'))), 'OneOf'('John')))) ;
%
%No
%==
%

owl_ace(OWL, ACE) :-
	ip(OWL, ACE, []).

%
% SYNTAX
%

ip(SubClassOf) --> nps(Y, SubClassOf), ibar(num=sg, Y), ['.'].

ibar(num=sg, C2) --> auxc(C1, C2), cop(C1).
ibar(Num, C2) --> auxv(Num, Neg, Part, C1, C2), vp(Num, Neg, Part, C1).

cop(C) --> [a], n(num=sg, C).
cop(C) --> propn(C).
cop('IntersectionOf'(C1, C2)) --> [a], n(num=sg, C1), relcoord(num=sg, C2).

vp(Num, Neg, Part, Restr) --> tv(Num, Neg, Part, R), npo(R, Restr).

nps(C, 'SubClassOf'(OneOf, C)) --> propn(OneOf).
%nps(C, 'ClassAssertion'(Individual, C)) --> propn('OneOf'(Individual)).

nps(Y, Out) --> dets(Num, X, Y, Out), n(Num, X).
nps(Y, Out) --> dets(Num, 'IntersectionOf'(X, Restr), Y, Out), n(Num, X), relcoord(Num, Restr).

npo(R, 'SomeValuesFrom'(R, OneOf)) --> propn(OneOf).
%npo(R, 'HasValue'(R, Individual)) --> propn('OneOf'(Individual)).

npo(R, 'ExistsSelf'(R)) --> [itself].
npo(R, Restr) --> detg(Num, R, C, Restr), n(Num, C).
npo(R, Restr) --> detg(Num, R, 'IntersectionOf'(C, RelCl), Restr), n(Num, C), relcoord(Num, RelCl).


relcoord(Num, Coord) --> relcoord_1(Num, X), relcoord_tail(Num, X, Coord).
relcoord_tail(Num, X, Coord) --> comma_and(X, Y, Coord), relcoord(Num, Y).
relcoord_tail(_, X, X) --> [].

relcoord_1(Num, Coord) --> relcoord_2(Num, X), relcoord_1_tail(Num, X, Coord).
relcoord_1_tail(Num, X, Coord) --> or(X, Y, Coord), relcoord_1(Num, Y).
relcoord_1_tail(_, X, X) --> [].

relcoord_2(Num, Coord) --> rel(Num, X), relcoord_2_tail(Num, X, Coord).
relcoord_2_tail(Num, X, Coord) --> and(X, Y, Coord), relcoord_2(Num, Y).
relcoord_2_tail(_, X, X) --> [].


rel(Num, X) --> relpro, ibar(Num, X).


%
% FORMAL LEXICON
%

dets(num=sg, C1, C2, 'SubClassOf'(C1, C2)) --> ['Every'].

% Syntactic sugar 'no' could be implemented like this:
% dets(num=sg, C1, C2, 'SubClassOf'(C1, 'ComplementOf'(C2))) --> ['No'].

detg(num=sg, R, C, 'SomeValuesFrom'(R, C)) --> [a].
detg(num=sg, R, C, 'AllValuesFrom'(R, C)) --> [nothing, but, a].
detg(num=pl, R, C, 'MinCardinality'(Number, R, C)) --> [at, least], positivenumber(Number).
detg(num=pl, R, C, 'MaxCardinality'(Number, R, C)) --> [at, most], positivenumber(Number).
detg(num=pl, R, C, 'ExactCardinality'(Number, R, C)) --> [exactly], positivenumber(Number).

auxc(C, C) --> [is].
auxc(C, 'ComplementOf'(C)) --> [is, not].

auxv(_, neg=no, vbn=no, C, C) --> [].
auxv(_, neg=no, vbn=yes, C, C) --> [is].
auxv(num=sg, neg=yes, vbn=no, C, 'ComplementOf'(C)) --> [does, not].
auxv(num=pl, neg=yes, vbn=no, C, 'ComplementOf'(C)) --> [do, not].
auxv(num=sg, neg=yes, vbn=yes, C, 'ComplementOf'(C)) --> [is, not].
auxv(num=pl, neg=yes, vbn=yes, C, 'ComplementOf'(C)) --> [are, not].

relpro --> [that].

comma_and(C1, C2, 'IntersectionOf'(C1, C2)) --> [',', and].
and(C1, C2, 'IntersectionOf'(C1, C2)) --> [and].
or(C1, C2, 'UnionOf'(C1, C2)) --> [or].

% General number rule (note that we don't support 0):
% positivenumber --> [PositiveNumber], { between(1, infinite, PositiveNumber) }.
positivenumber(2) --> [2].



%
% CONTENT LEXICON
%

% Propernames can come only in one form.
propn('OneOf'('John')) --> ['John'].

% Actually, propernames do not have to be included in the lexicon.
% We know from the context (i.e. missing determiner)
% what is a propername. Furthermore, propernames do not need
% morphological synthesis.
% propn('OneOf'(ProperName)) --> [ProperName].

% For nouns we have to describe 2 forms: {num=sg, num=pl}
n(num=sg, 'owl:Thing') --> [thing].
n(num=pl, 'owl:Thing') --> [things].
n(num=sg, man) --> [man].
n(num=pl, man) --> [men].

% For verbs we have to describe 8 forms:
% {neg=yes, neg=no} * {num=sg, num=pl} * {vbn=no, vbn=yes}

% likes
% like
% does not like
% do not like
% is liked by
% are liked by
% is not liked by
% are not liked by

% Note that the (external) lexicon has to provide only 3 forms:
% finite form, infinitive, and past participle (e.g. likes, like, liked).

tv(num=sg, neg=no, vbn=no, like) --> [likes].
tv(num=pl, neg=no, vbn=no, like) --> [like].
tv(num=sg, neg=yes, vbn=no, like) --> [like].
tv(num=pl, neg=yes, vbn=no, like) --> [like].
tv(_, _, vbn=yes, 'InverseObjectProperty'(like)) --> [liked, by].


%% pp
%
% Pritty printer. Generates all sentences allowed in the language.
% There is a restriction on the sentence length to get a finite output.
%
% Usage on the Unix commandline (where 'swipl' is a SWI-Prolog executable):
%
%==
% echo "[owlace_dcg]. owlace_dcg:pp." | swipl
%==

pp :-
	forall(
		( init_sentence(ACE), owl_ace(OWL, ACE) ),
		( concat_atom(ACE, ' ', ACEAtom), format('~w :: ~w~n', [ACEAtom, OWL]) )
	).

init_sentence(ACE) :-
	between(4, 12, Int),
	length(ACE, Int).