/**
* Assumes that the sentence expression lists have been generated.
*/
void prepare_to_parse(Sentence sent, Parse_Options opts)
{
size_t i;
build_sentence_disjuncts(sent, opts->disjunct_cost);
if (verbosity > 2) {
printf("After expanding expressions into disjuncts:");
print_disjunct_counts(sent);
}
print_time(opts, "Built disjuncts");
for (i=0; i<sent->length; i++) {
sent->word[i].d = eliminate_duplicate_disjuncts(sent->word[i].d);
/* Some long Russian sentences can really blow up, here. */
if (resources_exhausted(opts->resources))
return;
}
print_time(opts, "Eliminated duplicate disjuncts");
if (verbosity > 2) {
printf("\nAfter expression pruning and duplicate elimination:\n");
print_disjunct_counts(sent);
}
set_connector_length_limits(sent, opts);
pp_and_power_prune(sent, opts);
}
/**
* Increase the number of disjuncts associated to each word in the
* sentence by working with word-clusters. Return true if the number
* of disjuncts were expanded, else return false.
*/
bool lg_expand_disjunct_list(Sentence sent)
{
size_t w;
Cluster *clu = lg_cluster_new();
bool expanded = false;
for (w = 0; w < sent->length; w++)
{
X_node * x;
Disjunct * d = sent->word[w].d;
for (x = sent->word[w].x; x != NULL; x = x->next)
{
Disjunct *dx = build_expansion_disjuncts(clu, x);
if (dx)
{
unsigned int cnt = count_disjuncts(d);
d = catenate_disjuncts(dx, d);
d = eliminate_duplicate_disjuncts(d);
if (cnt < count_disjuncts(d)) expanded = true;
}
}
sent->word[w].d = d;
}
lg_cluster_delete(clu);
return expanded;
}
/**
* Assumes that the sentence expression lists have been generated.
*/
void prepare_to_parse(Sentence sent, Parse_Options opts)
{
size_t i;
build_sentence_disjuncts(sent, opts->disjunct_cost, opts);
if (verbosity_level(5))
{
prt_error("Debug: After expanding expressions into disjuncts:\n");
print_disjunct_counts(sent);
}
print_time(opts, "Built disjuncts");
for (i=0; i<sent->length; i++)
{
sent->word[i].d = eliminate_duplicate_disjuncts(sent->word[i].d);
/* Some long Russian sentences can really blow up, here. */
if (resources_exhausted(opts->resources))
return;
}
print_time(opts, "Eliminated duplicate disjuncts");
if (verbosity_level(5))
{
prt_error("Debug: After expression pruning and duplicate elimination:\n");
print_disjunct_counts(sent);
}
gword_record_in_connector(sent);
setup_connectors(sent);
}
void my_prepare_to_parse(Sentence sent, Parse_Options opts) {
/* assumes that the sentence expression lists have been generated */
/* this does all the necessary pruning and building of and */
/* structures. */
int i, has_conjunction;
// build_sentence_disjuncts(sent, opts->disjunct_cost);
// if (verbosity > 2) {
//printf("After expanding expressions into disjuncts:") ;
//print_disjunct_counts(sent);
// }
print_time(opts, "Built disjuncts");
for (i=0; i<sent->length; i++) {
sent->word[i].d = eliminate_duplicate_disjuncts(sent->word[i].d);
}
print_time(opts, "Eliminated duplicate disjuncts");
if (verbosity > 2) {
printf("\nAfter expression pruning and duplicate elimination:\n");
print_disjunct_counts(sent);
}
null_links = (opts->min_null_count > 0);
has_conjunction = sentence_contains_conjunction(sent);
set_connector_length_limits(sent, opts);
build_deletable(sent, has_conjunction);
build_effective_dist(sent, has_conjunction);
/* why do we do these here instead of in
first_prepare_to_parse() only? The
reason is that the deletable region
depends on if null links are in use.
with null_links everything is deletable */
if (!has_conjunction) {
pp_and_power_prune(sent, RUTHLESS, opts);
} else {
pp_and_power_prune(sent, GENTLE, opts);
/*if (verbosity > 2) {
printf("\nAfter Gentle power pruning:\n");
print_disjunct_counts(sent);
}
*/
/*print_time(opts, "Finished gentle power pruning"); */
conjunction_prune(sent, opts);
if (verbosity > 2) {
printf("\nAfter conjunction pruning:\n");
print_disjunct_counts(sent);
print_statistics();
}
print_time(opts, "Done conjunction pruning");
build_conjunction_tables(sent);
install_fat_connectors(sent);
install_special_conjunctive_connectors(sent);
if (verbosity > 2) {
printf("After conjunctions, disjuncts counts:\n");
print_disjunct_counts(sent);
}
set_connector_length_limits(sent, opts);
/* have to do this again cause of the
new fat connectors and disjuncts */
print_time(opts, "Constructed fat disjuncts");
prune(sent);
print_time(opts, "Pruned fat disjuncts");
for (i=0; i<sent->length; i++) {
sent->word[i].d = eliminate_duplicate_disjuncts(sent->word[i].d);
}
if (verbosity > 2) {
printf("After pruning and duplicate elimination:\n");
print_disjunct_counts(sent);
}
print_time(opts, "Eliminated duplicate disjuncts (again)");
if (verbosity > 2) print_AND_statistics(sent);
power_prune(sent, RUTHLESS, opts);
}
/*
if (verbosity > 2) {
printf("\nAfter RUTHLESS power-pruning:\n");
print_disjunct_counts(sent);
}
*/
/* print time for power pruning used to be here */
/* now done in power_prune itself */
print_time(opts, "Initialized fast matcher and hash table");
}