Merge pull request #1 from pzzhang/fix_pronoun_and_of

update parser and test examples

example/demo.py

@@ -37,6 +37,11 @@ def main():
  demo('The woman is a pianist.')
  demo('A giraffe grazing a tree in the wildness with other wildlife.')
  demo('Cow standing on sidewalk in city area near shops.')
+ demo('She is playing the piano.') # wrong, contain "she"
+ demo('A woman is next to a piano.') # wrong
+ demo('A woman is in front of a piano.') # wrong
+ demo('A woman is standing next to a piano.') # correct 
+ demo('A bowl of rice, noodle and grains.') # wrong
 
  print('Input your own sentence. Type q to quit.')
  while True:

sng_parser/backends/spacy_parser.py

@@ -74,7 +74,7 @@ def parse(self, sentence, doc=None, return_doc=False):
  entity_chunks = list()
  for entity in doc.noun_chunks:
  # Ignore pronouns such as "it".
- if entity.root.lemma_ == '-PRON-':
+ if entity.root.pos_ == "PRON":
  continue
 
  ent = dict(
@@ -174,14 +174,14 @@ def dfs(node):
  'relation': entity.root.head.text,
  'lemma_relation': entity.root.head.lemma_
  }
- # E.g., A [woman] in front of a [piano].
+ # E.g., A [woman] (is) in front of a [piano].
  elif (
  entity.root.head.head.dep_ == 'pobj' and
  database.is_phrasal_prep(doc[entity.root.head.head.head.i:entity.root.head.i + 1].text.lower())
  ):
  fake_noun_marks.add(entity.root.head.head.i)
  relation = {
- 'subject': entity.root.head.head.head.head.i,
+ 'subject': relation_subj[entity.root.head.head.head.head.i] if entity.root.head.head.head.head.pos_ == 'AUX' else entity.root.head.head.head.head.i,
  'object': entity.root.i,
  'relation': doc[entity.root.head.head.head.i:entity.root.head.i + 1].text,
  'lemma_relation': doc[entity.root.head.head.head.i:entity.root.head.i].lemma_
@@ -202,6 +202,14 @@ def dfs(node):
  'relation': entity.root.head.head.text + ' ' + entity.root.head.text,
  'lemma_relation': entity.root.head.head.lemma_ + ' ' + entity.root.head.lemma_
  }
+ # E.g., A [piano] is next to a [woman].
+ elif entity.root.head.head.dep_ == 'acomp' and entity.root.head.head.head.pos_ == 'AUX':
+ relation = {
+ 'subject': relation_subj[entity.root.head.head.head.i],
+ 'object': entity.root.i,
+ 'relation': entity.root.head.head.text + ' ' + entity.root.head.text,
+ 'lemma_relation': entity.root.head.head.lemma_ + ' ' + entity.root.head.lemma_
+ }
  # E.g., A [woman] standing next to a [piano].
  elif entity.root.head.head.dep_ in ('amod', 'advmod') and entity.root.head.head.head.pos_ == 'VERB' and entity.root.head.head.head.i in relation_subj:
  relation = {
@@ -272,5 +280,5 @@ def __flatten_conjunction(node):
  yield node
  for c in node.children:
  if c.dep_ == 'conj':
- yield c
+ yield from SpacyParser.__flatten_conjunction(c)