Prolog递归语法

时间:2012-05-01 15:14:12

标签: prolog dcg

目前我遇到了从noun_phrase循环回np2的问题。我想知道是否有人可以帮助我回到noun_phrase。这是一些代码:

noun_phrase([X|T],(det(X), NP2),Rem):-
   det(X),
   np2(T,NP2,Rem).

np2([H|T],np2(adj(H),Rest),NP) :-
   adj(H),
   np2(T,Rest,Rem),
   noun_phrase(NP,Rem,_).

我想从np2循环回noun_phrase。我认为np2的代码是错误的,因为我只是一起攻击它。

1 个答案:

答案 0 :(得分:10)

直接在Prolog中编码语法是一个非常麻烦的过程。是的,你可以这样做,但如果你刚开始学习Prolog,你就不是最好的位置。事实上,科学花了很多年才提出一种特别有效的编码 - 当这种编码被理解时,Prolog就诞生了!

您需要的是语法 - definite clause grammars 。不要试图在Prolog(现在)中理解他们的编码,只需习惯phrase/2

以下是来自Prolog and Natural-Language Analysis的程序3.11 Fernando C. N. Pereira和Stuart M. Shieber。 Pereira设计了我们今天使用的DCG规则。这本书是最被低估的Prologbooks之一。它是免费的!

s(s(NP,VP)) --> np(NP), vp(VP).

np(np(Det,N,Rel)) -->
   det(Det),
   n(N),
   optrel(Rel).
np(np(PN)) --> pn(PN).

vp(vp(TV,NP)) --> tv(TV), np(NP).
vp(vp(IV)) --> iv(IV).

optrel(rel(epsilon)) --> [].
optrel(rel(that,VP)) --> [that], vp(VP).

pn(pn(terry)) --> [terry].
pn(pn(shrdlu)) --> [shrdlu].

iv(iv(halts)) --> [halts].

det(det(a)) --> [a].

n(n(program)) --> [program].

tv(tv(writes)) --> [writes].

如果您想将词典表示为Prolog事实,请使用

代替
n(n(program)) --> [program].

,而

n(n(W)) --> [W],{noun(W)}.

noun(program).

所以,让我们使用它:

?- phrase(s(P), Xs).
P = s(np(det(a),n(program),rel(epsilon)),vp(tv(writes),np(det(a),n(program),rel(epsilon)))),
Xs = [a,program,writes,a,program] ;
P = s(np(det(a),n(program),rel(epsilon)),vp(tv(writes),np(det(a),n(program),rel(that,vp(tv(writes),np(det(a),n(program),rel(epsilon))))))),
Xs = [a,program,writes,a,program,that,writes,a,program] ...

多么自我指涉的话语!但是现在,所有句子的长度都更具启发性:

?- length(Xs, N), phrase(s(P), Xs).
Xs = [terry,halts],
N = 2,
P = s(np(pn(terry)),vp(iv(halts))) ;
Xs = [shrdlu,halts],
N = 2,
P = s(np(pn(shrdlu)),vp(iv(halts))) ;
Xs = [a,program,halts],
N = 3,
P = s(np(det(a),n(program),rel(epsilon)),vp(iv(halts))) ;
Xs = [terry,writes,terry],
N = 3,
P = s(np(pn(terry)),vp(tv(writes),np(pn(terry)))) ;
Xs = [terry,writes,shrdlu],
N = 3,
P = s(np(pn(terry)),vp(tv(writes),np(pn(shrdlu)))) ;
Xs = [shrdlu,writes,terry],
N = 3,
P = s(np(pn(shrdlu)),vp(tv(writes),np(pn(terry)))) ...