/**
* This function generates a table, word_used[i][w], showing
* whether each word w is used in each sublinkage i; if so,
* the value for that cell of the table is 1.
*/
static void count_words_used(con_context_t *ctxt, Linkage linkage)
{
int i, w, link, num_subl;
num_subl = linkage->num_sublinkages;
if(linkage->unionized==1 && num_subl>1) num_subl--;
if (verbosity>=2)
printf("Number of sublinkages = %d\n", num_subl);
for (i=0; i<num_subl; i++) {
for (w=0; w<linkage->num_words; w++) ctxt->word_used[i][w]=0;
linkage->current=i;
for (link=0; link<linkage_get_num_links(linkage); link++) {
ctxt->word_used[i][linkage_get_link_lword(linkage, link)]=1;
ctxt->word_used[i][linkage_get_link_rword(linkage, link)]=1;
}
if (verbosity>=2) {
printf("Sublinkage %d: ", i);
for (w=0; w<linkage->num_words; w++) {
if (ctxt->word_used[i][w]==0) printf("0 ");
if (ctxt->word_used[i][w]==1) printf("1 ");
}
printf("\n");
}
}
}
/**
* Go through all the words. If a word is on the right end of
* an S (or SF or SX), wordtype[w]=STYPE. If it's also on the left end of a
* Pg*b, I, PP, or Pv, wordtype[w]=PTYPE. If it's a question-word
* used in an indirect question, wordtype[w]=QTYPE. If it's a
* question-word determiner, wordtype[w]=QDTYPE. Else wordtype[w]=NONE.
* (This function is called once for each sublinkage.)
*/
static void generate_misc_word_info(con_context_t * ctxt, Linkage linkage)
{
int l1, l2, w1, w2;
const char * label1, * label2;
for (w1=0; w1<linkage->num_words; w1++)
ctxt->wordtype[w1]=NONE;
for (l1=0; l1<linkage_get_num_links(linkage); l1++) {
w1=linkage_get_link_rword(linkage, l1);
label1 = linkage_get_link_label(linkage, l1);
if ((uppercompare(label1, "S")==0) ||
(uppercompare(label1, "SX")==0) ||
(uppercompare(label1, "SF")==0)) {
ctxt->wordtype[w1] = STYPE;
for (l2=0; l2<linkage_get_num_links(linkage); l2++) {
w2=linkage_get_link_lword(linkage, l2);
label2 = linkage_get_link_label(linkage, l2);
if ((w1==w2) &&
((post_process_match("Pg#b", label2)==1) ||
(uppercompare(label2, "I")==0) ||
(uppercompare(label2, "PP")==0) ||
(post_process_match("Pv", label2)==1))) {
/* Pvf, Pgf? */
ctxt->wordtype[w1] = PTYPE;
}
}
}
if (post_process_match("QI#d", label1)==1) {
ctxt->wordtype[w1] = QTYPE;
for (l2=0; l2<linkage_get_num_links(linkage); l2++) {
w2=linkage_get_link_lword(linkage, l2);
label2 = linkage_get_link_label(linkage, l2);
if ((w1==w2) && (post_process_match("D##w", label2)==1)) {
ctxt->wordtype[w1] = QDTYPE;
}
}
}
if (post_process_match("Mr", label1)==1) ctxt->wordtype[w1] = QDTYPE;
if (post_process_match("MX#d", label1)==1) ctxt->wordtype[w1] = QDTYPE;
}
}
/// This is the basic sentence dissection
static PyObject *sentence(PyObject *self, PyObject *args) {
Dictionary dict;
Parse_Options opts;
Sentence sent;
Linkage linkage;
Linkage sub_linkage;
char * diagram;
/// Link counts
int num_linkages;
int links;
/// Index's for the iterators
int link_idx;
int word_idx;
int num_words;
long span;
long sub_linkages;
const char *text;
const char *d_output;
PyObject *output_list;
PyObject *word_list;
PyObject *word2_list;
PyObject *span_list;
PyObject *temp;
PyObject *sublinkage_list;
PyObject *_diagram;
output_list = PyList_New(0);
word_list = PyList_New(0);
word2_list = PyList_New(0);
sublinkage_list = PyList_New(0);
span_list = PyList_New(0);
if (!PyArg_ParseTuple(args, "s", &text))
return NULL;
opts = parse_options_create();
parse_options_set_verbosity(opts, -1);
parse_options_set_screen_width(opts, 50);
setlocale(LC_ALL, "");
dict = dictionary_create_default_lang();
if (!dict) {
PyErr_SetString(PyExc_RuntimeError, "Fatal error: Unable to open the dictionary");
Py_INCREF(Py_None);
return Py_None;
}
sent = sentence_create(text, dict);
sentence_split(sent, opts);
num_linkages = sentence_parse(sent, opts);
if (num_linkages > 0) {
linkage = linkage_create(0, sent, opts);
/// Get the lengths of everything
num_words = linkage_get_num_words(linkage);
links = linkage_get_num_links(linkage);
for(link_idx=0; link_idx < links; link_idx++) {
PyObject *temp_subLen;
diagram = linkage_print_diagram(linkage);
_diagram = PyString_FromString(diagram);
sub_linkage = linkage_create(link_idx, sent, opts);
sub_linkages = linkage_get_num_sublinkages(linkage);
temp_subLen = PyLong_FromLong(sub_linkages);
linkage_delete(sub_linkage);
PyList_Append(sublinkage_list, temp_subLen);
span = linkage_get_link_length(linkage, link_idx);
PyList_Append(span_list, PyInt_FromLong(span));
PyObject *temp_list;
temp_list = PyList_New(0);
/// Sub Group these (left and right labels)
const char *t1 = linkage_get_link_llabel(linkage, link_idx);
temp = PyString_FromString(t1);
PyList_Append(temp_list, temp);
const char *t2 = linkage_get_link_rlabel(linkage, link_idx);
temp = PyString_FromString(t2);
PyList_Append(temp_list, temp);
/// Then add to the main list
PyList_Append(output_list, temp_list);
/// Just the label
const char *t3 = linkage_get_link_label(linkage, link_idx);
temp = PyString_FromString(t3);
PyList_Append(word2_list, temp);
}
for(word_idx=0; word_idx < num_words; word_idx++) {
d_output = linkage_get_word(linkage, word_idx);
PyObject *word;
word = PyString_FromString(d_output);
PyList_Append(word_list, word);
}
linkage_free_diagram(diagram);
linkage_delete(linkage);
} else {
sentence_delete(sent);
dictionary_delete(dict);
parse_options_delete(opts);
Py_INCREF(Py_None);
return Py_None;
}
sentence_delete(sent);
dictionary_delete(dict);
parse_options_delete(opts);
return Py_BuildValue("SSSSSS", word_list, span_list, output_list, word2_list, sublinkage_list, _diagram);
}
VALUE linkage_count_num_links(const VALUE self, VALUE link) {
LinkagePtr *link_ptr = retrieve_linkage(link);
int cnt = linkage_get_num_links(link_ptr->linkage);
return INT2FIX(cnt);
}
/*
* Class: LinkGrammar
* Method: getNumLinks
* Signature: ()I
*/
JNIEXPORT jint JNICALL
Java_org_linkgrammar_LinkGrammar_getNumLinks(JNIEnv *env, jclass cls)
{
per_thread_data *ptd = get_ptd(env, cls);
return linkage_get_num_links(ptd->linkage);
}
SPOTriplets NLP::sentence2triplets ( const char* sentence )
{
// vector of triplets
SPOTriplets triplets;
#ifdef DEBUG
std::cout << "The sentence: " << sentence << std::endl;
#endif
// creates a Sentence from the input char*
Sentence sent = sentence_create ( sentence, dict_ );
#ifdef DEBUG
std::cout << "Sentence created" << std::endl;
#endif
// tokenizes the sentence
sentence_split ( sent, parse_opts_ );
#ifdef DEBUG
std::cout << "Sentence splitted" << std::endl;
#endif
// searches for all possible linkages
int num_linkages = sentence_parse ( sent, parse_opts_ );
#ifdef DEBUG
std::cout << "Sentence parsed" << std::endl;
std::cout << "Number of linkages: " << num_linkages << std::endl;
#endif
// just one triplet
SPOTriplet triplet;
// if there is any linkage in the sentence
if( num_linkages > 0 )
{
// create the linkage
Linkage linkage = linkage_create ( 0, sent, parse_opts_ );
#ifdef DEBUG
// prints the sentence's diagram
std::cout << "The diagram: " << std::endl;
char *diagram = linkage_print_diagram(linkage, true, 800);
std::cout << diagram << std::endl;
linkage_free_diagram( diagram );
// end print diagram
#endif
std::vector<std::string> labels;
// 1. find the S_link
// S* except there is an SJ* because then S* except Spx
// two cases: there is SJ* and there is not SJ*
// TODO: VJlp VJrp same as SJ but to predications
// TODO: SFut SFst what the f**k? ###FIXED###
// TODO: His form was shining like the light not working ###FIXED###
// TODO: Car is mine not working ###FIXED###
// TODO: The little brown bear has eaten all of the honey not working ###FIXED###
// REGEXES
std::regex SJ_( "SJ.*" );
std::regex VJ_( "VJ.*");
std::regex subject( "(Ss.*)|(SFut)|(Sp\*.*)" );
std::regex Spx( "Spx.*" );
// TODO:fix theese initializer list not allowed ###FIXED###
std::regex predicate( "(Pv.*)|(Pg.*)|(PP.*)|(I.*)|(TO)|(MVi.*)" );
// TODO: make one from theese // (Sp.*)|(Ss.*) ###FIXED###
std::regex noun_adject_object ( "(O.*)|(Os.*)|(Op.*)|(MVpn.*)|(Pa.*)|(MVa.*)" );
std::regex preposition ( "(MVp.*)|(Pp.*)|(OF)|(TO)" );
std::regex prep_object ( "(J.*)|(TI)|(I.*)|(ON)" );
// TODO: problems with matching!! Pg*!! ###FIXED###
// TODO: problems with matching!! Mvp.*!! ###FIXED###
bool s_found = false;
bool p_found = false;
bool o_found = false;
bool SJ = false;
// search for SJ.s labels
for( auto label: labels )
{
if( std::regex_match( label, SJ_ ) )
{
SJ = true;
break;
}
}
// multiple subject in the sentence
if( SJ )
{
// SPls left -> first subject
// SPrs right -> second subject
// Spx right -> predicate
// SJ-s are multiple subjects
std::string temp;
// go through every linkage
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
// get their label
std::string l = linkage_get_link_label( linkage, i );
// if there is an SJl* label
if( std::regex_match( l, std::regex( "SJl.*" ) ) )
{
// SJls left side
triplet.s = linkage_get_word( linkage, linkage_get_link_lword( linkage, i ) );
triplet.cut( triplet.s );
temp = triplet.s + " ";
// and word
triplet.s = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
triplet.cut( triplet.s );
temp += triplet.s + " ";
// find SJr*
for( int j = 0; j < linkage_get_num_links( linkage ); ++j )
{
std::string m = linkage_get_link_label( linkage, j );
if( std::regex_match( m, std::regex( "SJr.*" ) ) )
{
triplet.s = linkage_get_word( linkage, linkage_get_link_rword( linkage, j ) );
triplet.cut();
temp += triplet.s;
triplet.s = temp;
s_found = true;
#ifdef DEBUG
std::cout << "Subject found: " << triplet.s << std::endl;
#endif
break;
} // if
} // for
break;
} // if
} // for
// now we have the subject
// find Spx and its right side will be the starter predicate
std::string current_word;
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
std::string l = linkage_get_link_label( linkage, i );
if( std::regex_match( l, std::regex( "Spx.*" ) ) )
{
triplet.p = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
current_word = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
}
}
// from now all the same as on the else branch !!!!
bool predicate_match = false;
// search for the linkage that has triplet.s as left!
do
{
predicate_match = false;
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
// every linkage's left word
std::string word_i = linkage_get_word( linkage, linkage_get_link_lword( linkage, i ) );
// every linkage's label
std::string l = linkage_get_link_label( linkage, i );
if( std::regex_match( l, predicate ) && word_i == current_word )
{
// found predicate
triplet.p = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
current_word = triplet.p;
predicate_match = true;
break;
}
}
}
while( predicate_match );
// we now have the predicate too
// TODO: multiple predicates!
p_found = true;
#ifdef DEBUG
std::cout << "Predicate found: " << triplet.p << std::endl;
#endif
// ###COPY BEGIN###
// search for noun object or adjective object
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
// get every linkage label
std::string l = linkage_get_link_label( linkage, i );
// get the left word of every linkage
std::string l_word = linkage_get_word( linkage, linkage_get_link_lword( linkage, i ) );
// if thete is a label that match AND its left word is the predicate
if( std::regex_match( l, noun_adject_object ) && triplet.p == l_word )
{
// then the object is that linkage's right word
triplet.o = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
triplet.cut( triplet.o );
o_found = true;
#ifdef DEBUG
std::cout << "Adjective or noun object found: " << triplet.o << std::endl;
#endif
} // if
} // for
// still not found object, then search for preposition
if( !o_found )
{
// go through every linkage
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
// get the linkage's label
std::string l = linkage_get_link_label( linkage, i );
// and left word
std::string word_i = linkage_get_word( linkage, linkage_get_link_lword( linkage, i ) );
// if there is a linkage which is a preposition and its left word is the predicate
if( std::regex_match( l, preposition ) && triplet.p == word_i )
{
// found preposition
// search for prep_object
// then the temp will contain the preposition label's right word
std::string temp = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
#ifdef DEBUG
std::cout << "Preposition found! and its rigth word is: " << temp << std::endl;
#endif
for( int j = 0; j < linkage_get_num_links( linkage ); ++j )
{
// every linkages
std::string m = linkage_get_link_label( linkage, j );
// every left word
std::string word_j = linkage_get_word( linkage, linkage_get_link_lword( linkage, j ) );
// if there is a label with match and its left is exactly the preposition's right
if( std::regex_match( m, prep_object ) && temp == word_j )
{
triplet.o = linkage_get_word( linkage, linkage_get_link_lword( linkage, j ) );
triplet.cut(triplet.o);
triplet.o += " ";
// save o
std::string temp = triplet.o;
triplet.o = linkage_get_word( linkage, linkage_get_link_rword( linkage, j ) );
triplet.cut(triplet.o);
temp += triplet.o;
triplet.o = temp;
o_found = true;
#ifdef DEBUG
std::cout << "Object found: " << triplet.o << std::endl;
#endif
} // if( std::regex_match( m, prep_object ) && temp == word_j ) END
} // for J END
} // if( std::regex_match( l, preposition ) && triplet.p == word_i ) END
} // for I END
} // if( !o_found ) END
if( s_found && p_found && o_found )
{
// TODO: cut the words itself not the whole triplet
// have to cut every word itself
// triplet.cut();
triplet.cut(triplet.s);
triplet.cut(triplet.p);
triplets.push_back( triplet );
s_found = false;
p_found = false;
o_found = false;
}
// ###COPY END###
}
else // only one subject
{
// except Spx!!!
// S left -> subject
// S right -> predicate at first
// if the word next to S right, is an element of Pv*, Pg* PP*, I*, TO, MVi*
// then the new predicate will be that word
std::string current_word;
// search for subject (S_link)
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
// get the linkage's label
std::string l = linkage_get_link_label( linkage, i );
if( std::regex_match( l, subject ) )
{
// subject found
triplet.s = linkage_get_word( linkage, linkage_get_link_lword( linkage, i ) );
s_found = true;
current_word = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
triplet.p = current_word;
#ifdef DEBUG
std::cout << "Subject found: " << triplet.s << std::endl;
#endif
break;
}
}
if( s_found )
{
bool predicate_match = false;
// search for the linkage that has triplet.s as left!
do
{
predicate_match = false;
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
// every linkage's left word
std::string l_word = linkage_get_word( linkage, linkage_get_link_lword( linkage, i ) );
// every linkage's label
std::string l = linkage_get_link_label( linkage, i );
if( std::regex_match( l, predicate ) && l_word == current_word )
{
// found predicate
triplet.p = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
current_word = triplet.p;
predicate_match = true;
break;
}
} // for END
} while( predicate_match );
p_found = true;
#ifdef DEBUG
std::cout << "Predicate found: " << triplet.p << std::endl;
#endif
} // if( s_found ) END
// subject and predicate found
// search for object
// from k to linkage_get_num_links( linkage )
// if there is any of the noun, adjective od preposition object then that
// label's right will give the object.
// !!! search only between labels that has triplet.p as left word !!!!!
// search for noun object or adjective objects
// go through all links
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
// get every linkage label
std::string l = linkage_get_link_label( linkage, i );
// get the left word of every linkage
std::string word_i = linkage_get_word( linkage, linkage_get_link_lword( linkage, i ) );
// if thete is a label that match AND its left word is the predicate
if( std::regex_match( l, noun_adject_object ) && triplet.p == word_i )
{
// then the object is that linkage's right word
triplet.o = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
o_found = true;
triplet.cut(triplet.o);
#ifdef DEBUG
std::cout << "Adjective or noun object found: " << triplet.o << std::endl;
#endif
} // if END
} // for END
// still not found object, then search for preposition
if( !o_found )
{
// go through every linkage
for( int i = 0; i < linkage_get_num_links( linkage ); ++i )
{
// get the linkage's label
std::string l = linkage_get_link_label( linkage, i );
// and left word
std::string word_i = linkage_get_word( linkage, linkage_get_link_lword( linkage, i ) );
// if there is a linkage which is a preposition and its left word is the predicate
if( std::regex_match( l, preposition ) && triplet.p == word_i )
{
// found preposition
// search for prep_object
// then the temp will contain the preposition label's right word
std::string temp = linkage_get_word( linkage, linkage_get_link_rword( linkage, i ) );
#ifdef DEBUG
std::cout << "Preposition found! and its rigth word is: " << temp << std::endl;
#endif
// start search from there
for( int j = 0; j < linkage_get_num_links( linkage ); ++j )
{
// every linkages
std::string m = linkage_get_link_label( linkage, j );
// every left word
std::string word_j = linkage_get_word( linkage, linkage_get_link_lword( linkage, j ) );
#ifdef DEBUG
if( std::regex_match( m, prep_object ) )
std::cout << m << " DOES match to (J.*)|(TI)|(I.*)|(ON)" << std::endl;
#endif
// if there is a label with match and its left is exactly the preposition's right
if( std::regex_match( m, prep_object ) && temp == word_j )
{
triplet.o = linkage_get_word( linkage, linkage_get_link_lword( linkage, j ) );
triplet.cut(triplet.o);
triplet.o += " ";
// save o
std::string temp = triplet.o;
triplet.o = linkage_get_word( linkage, linkage_get_link_rword( linkage, j ) );
triplet.cut(triplet.o);
temp += triplet.o;
triplet.o = temp;
#ifdef DEBUG
std::cout << "Object found: " << triplet.o << std::endl;
#endif
o_found = true;
}
} // for
} // if
} // for
} // if( o_found ) END
if( s_found && p_found && o_found )
{
// TODO: cut the words itself not the whole triplet ###FIXED###
// have to cut every word itself
// triplet.cut();
triplet.cut(triplet.s);
triplet.cut(triplet.p);
triplets.push_back( triplet );
s_found = false;
p_found = false;
o_found = false;
}
} // end else
linkage_delete ( linkage );
} // if( num_linkages > 0 ) END