def _get_prep_phrases(self, head):
prep_phrases = []
prep_list = self._get_dependents(self._dependencies['prep'], head)
if prep_list:
for prep in prep_list:
if prep.word.lower() not in self._prep_blacklist_for_prep_phrases:
# Ignore those prepositions that are far away from the head
if abs(prep.index - head.index) < 3:
# Look for pobj
obj_list = self._get_dependents(self._dependencies['pobj'], prep)
if obj_list:
for obj in obj_list:
if not self._head_extracting_condition(obj, pos=True):
continue
obj_conjunction = self._get_conjunction(obj)
for o in obj_conjunction:
obj_seq = self._expand_head_word(o)
if obj_seq:
obj_seq.add_word_unit(prep)
prep_phrase = WordUnitSequence()
prep_phrase.extend(obj_seq)
prep_phrase.head = o
prep_phrase.nn_head = obj_seq.nn_head
if prep_phrase:
self._print_expansion_debug_info(head, 'prep phrase', prep_phrase)
prep_phrases.append(prep_phrase)
return prep_phrases
def _get_neg_modifier(self, head):
neg_mod = WordUnitSequence()
neg_list = self._get_dependents(self._dependencies['neg'], head)
if neg_list and neg_list[0].pos == self._pos_tags['dt']:
neg_mod.add_word_unit(neg_list[0])
self._print_expansion_debug_info(head, 'negation', neg_list[0])
return neg_mod
def _get_num_modifier(self, head):
num_mod = WordUnitSequence()
num_list = self._get_dependents(self._dependencies['num'], head)
if num_list:
for num in num_list:
num_mod.add_word_unit(num)
self._print_expansion_debug_info(head, 'numeric modifier', num)
return num_mod
def _get_noun_compound(self, head):
nc = WordUnitSequence()
nn_list = self._get_dependents(self._dependencies['nn'], head)
if nn_list:
for nn in nn_list:
nc.add_word_unit(nn)
self._print_expansion_debug_info(head, 'noun compound', nn)
return nc
def _get_vmod_phrase(self, head):
vmod_phrase = WordUnitSequence()
vmod_list = self._get_dependents(self._dependencies['vmod'], head)
if vmod_list:
for vmod in vmod_list:
predicate_object = self._get_predicate_object(vmod)
if predicate_object:
predicate, object = predicate_object[0]
vmod_phrase.extend(predicate)
vmod_phrase.extend(object)
self._print_expansion_debug_info(head, 'vmod', vmod_phrase)
return vmod_phrase
def _get_predicate_object(self, pred_head):
predicate_object = []
predicates = self._expand_predicate(pred_head)
for predicate in predicates:
dobj_flag, acomp_flag, pobj_flag = False, False, False
for ind, pred in predicate:
# Look for direct object
obj_list = self._get_dependents(self._dependencies['dobj'], pred)
if obj_list:
for obj in obj_list:
if not self._head_extracting_condition(obj, pos=True):
continue
obj_conjunction = self._get_conjunction(obj)
for o in obj_conjunction:
expanded_obj = self._expand_head_word(o)
if expanded_obj:
object = WordUnitSequence()
object.extend(expanded_obj)
object.head = o
object.nn_head = expanded_obj.nn_head
dobj_flag = True
predicate_object.append((predicate, object))
continue
# Look for adjective compliment
acomp_list = self._get_dependents(self._dependencies['acomp'], pred)
if acomp_list:
for acomp in acomp_list:
acomp_prep_phrases = self._get_prep_phrases(acomp)
for acomp_prep_phrase in acomp_prep_phrases:
if len(acomp_prep_phrase) > 1:
object = WordUnitSequence()
object.extend(WordUnitSequence(acomp_prep_phrase[1:]))
object.head = acomp_prep_phrase.head
object.nn_head = acomp_prep_phrase.nn_head
# Make a copy of predicate in case it gets expanded
predicate_copy = deepcopy(predicate)
# Merge the acomp and prep into the predicate
predicate_copy.add_word_unit(acomp)
predicate_copy.add_word_unit(acomp_prep_phrase[0])
acomp_flag = True
predicate_object.append((predicate_copy, object))
continue
# Look for prepositional objects
prep_phrases = self._get_prep_phrases(pred)
for prep_phrase in prep_phrases:
if len(prep_phrase) > 1:
object = WordUnitSequence()
object.extend(WordUnitSequence(prep_phrase[1:]))
object.head = prep_phrase.head
object.nn_head = prep_phrase.nn_head
# Make a copy of predicate in case it gets expanded
predicate_copy = deepcopy(predicate)
# Merge the prep into the predicate
predicate_copy.add_word_unit(prep_phrase[0])
pobj_flag = True
predicate_object.append((predicate_copy, object))
# Also return the predicate if it has no object in case it is a conjunction of other predicates.
if not dobj_flag and not acomp_flag and not pobj_flag:
predicate_object.append((predicate, None))
return predicate_object
def _expand_head_word(self, head):
def _clean(word_unit_seq):
# If the sequence is a single letter, ignore it
if len(word_unit_seq) == 1 and len(word_unit_seq[0]) == 1:
word_unit_seq = None
# If the head of the sequence is a number, ignore it
if word_unit_seq.head.pos == self._pos_tags['cd']:
word_unit_seq = None
return word_unit_seq
expansion = WordUnitSequence(head, head)
# Find out if the head is in a compound noun
noun_compound = self._get_noun_compound(head)
expansion.extend(noun_compound)
expansion.nn_head = deepcopy(expansion)
# # Find out if there is any numeric modifier
# num_mod = self._get_num_modifier(head)
# expansion.extend(num_mod)
# Find out if there is any negation
neg_mod = self._get_neg_modifier(head)
expansion.extend(neg_mod)
# Find out if the head has pobj phrase
pobj_phrases = self._get_prep_phrases(head)
if pobj_phrases:
pobj_phrase = pobj_phrases[0]
expansion.extend(pobj_phrase)
# Transfer the head in the pattern "<num> of <noun>" from "<num>" to "<noun>"
if (head.pos == self._pos_tags['cd'] or head.word.isdigit()) and pobj_phrase[0].word == 'of':
expansion.head = pobj_phrase.head
expansion.nn_head = pobj_phrase.nn_head
# Find out if the head has vmod phrase
vmod_phrase = self._get_vmod_phrase(head)
expansion.extend(vmod_phrase)
# Cleaning
expansion = _clean(expansion)
return expansion