Extremely inefficient and unelegant code

Question

I have a problem with my Latin translator which I am building for a school project. I have created a grammar for Latin/English sentences. I have created a predicate (possiblesentence(Latin,English).) which can translate a Latin sentence in the right order. However, as Latin sentences are in any order, I have resorted to generating a big list of permutations to find the right order of Latin to translate into English. This is my translate(Latin,English) predicate. This seems very inefficient. Is there a better way of doing this?

I also have another file that loads a big set of predicates from a .csv that look like this:

noun("femin","femin",1,f,"woman","women").
verb("port","porta","portav","portat", 1, carry, carrying, carried).

Here is my main program:

% Nick's Latin translator

% --------------
% word generator
% --------------

% predicate used to see if any two of the stems join to the ending.

stem_ending_joiner(Stem1,Stem2,Ending,Latin) :-
    (
        append(Stem1,Ending,Latin),
        !
    ;   append(Stem2,Ending,Latin),
        !
    ).

% predicate containing all possible data about nouns including endings
% and such

word_nounx(Latin,Translation,Gender,Case,Number) :-
    noun(Stem1,Stem2,_,Gender,Sgtrans,Pltrans),
    nounending(Ending,Prefix,Number,Case),
    stem_ending_joiner(Stem1,Stem2,Ending,Latin),
    (
        Number = sg,
        Translationw = Sgtrans;
        Number = pl,
        Translationw = Pltrans
    ),
    append(Prefix," ",Prefixspace),
    append(Prefixspace,Translationw,Translation).

% predicate containing all possible data about verbs including endings
% and such
% arguments: Latin,Translation,Tense,Number,Mood,Voice

word_verbx(Latin,Translation,Tense,Number,Mood,Voice,Person) :-
    verb(Stem1,_,_,_,_,Ptrans,_,Pastrans),
    verbending(Ending,Tense,Prefix,Number,Mood,Voice,Person),
    stem_ending_joiner(Stem1,_,Ending,Latin),
    (
        Tense = present,
        Translationw = Ptrans;
        Tense = past,
        Translationw = Pastrans
    ),
    append(Prefix," ",Prefixspace),
    append(Prefixspace,Translationw,Translation).

% -------
% parsing
% -------

word_noun(Latin,Translation,Gender,Case,Number) :-
    word_nounx(Latinx,Translationx,Gender,Case,Number),
    name(Latin,Latinx),
    name(Translation,Translationx).

word_verb(Latin,Translation,Tense,Number,Mood,Voice,Person) :-
    word_verbx(Latinx,Translationx,Tense,Number,Mood,Voice,Person),
    name(Latin,Latinx),
    name(Translation, Translationx).

pnoun(Gender,Case,Number) -->
   [[English,Latin]],
   {word_noun(Latin,English,Gender,Case,Number)}.
nounphrase(Gender,Case,Number) -->
   pnoun(Gender,Case,Number).
pverb(Tense,Number,Mood,Voice,Person) -->
   [[English,Latin]],
   {word_verb(Latin,English,Tense,Number,Mood,Voice,Person)}.
verbphrase(Tense,Number,Mood,Voice,Person,Gender,Case,Nnumber) -->
   pverb(Tense,Number,Mood,Voice,Person),
   nounphrase(Gender,Case,Nnumber).
sentence -->
   nounphrase(_,nom,Number),
   verbphrase(_,Number,_,_,3,_,nom,_).

% Predicates which manipulate lists of list-pairs to get the first and
% last elements of each list.

headofelements([],[]).
headofelements([H|T],[[H|_]|T1]) :-
    headofelements(T,T1).

tailofelements([],[]).
tailofelements([H|T],[[_|H]|T1]) :-
    tailofelements(T,T1).

lastofelements([],[]).
lastofelements([H|T],[X|T1]) :-
    last(X,N),
    H = N,
    lastofelements(T,T1).

% generates possible sentences with the latin and the english.

possible_sentence(X,N) :-
    phrase(sentence,Y),
    lastofelements(X,Y),
    headofelements(N,Y).

translate(Latin,English) :-
    permutation(Latin,X),
    possible_sentence(X,English).

Here are the endings:

% --------------------------
% ADJECTIVE ENDINGS
% ------------------------------

adjending("us",sg,nom,m).

% -----------------------------------
% VERBENDINGS
% ------------------------------------

verbending("o",present,"I",sg,_,_,1).
verbending("as",present,"you",sg,_,_,2).
verbending("at",present,"He",sg,_,_,3).
verbending("amus",present,"We",pl,_,_,1).
verbending("atis",present,"You",pl,_,_,2).
verbending("ant",present,"They",pl,_,_,3).


% -----------------------------------
% NOUN ENDINGS
% -----------------------------------

nounending("a","",sg,nom).
nounending("am","",sg,acc).
nounending("ae","of",sg,gen).
nounending("ae","to",sg,dat).
nounending("a","with",sg,abl).
nounending("ae","",pl,nom).
nounending("as","",pl,acc).
nounending("arum","of",pl,gen).
nounending("is","to",pl,dat).
nounending("is","with",pl,abl).

Answer 1

Latin should be a favourite application for Free-Word-Order Dependency Parsing.

The ProNTo Toolkit describes such technique and provides some introductory SW.

Incidentally, the dependency parser documentation uses Latin as primary application domain.