def appositive(pair,parsed_sentences):
if pair.first.sentenceID == pair.second.sentenceID:
index1 = eval(pair.first.offsets[0])
index2 = eval(pair.second.offsets[0])
#print index1,index2
#print pair.first.str,pair.second.str
tree = parsed_sentences[eval(pair.first.sentenceID)]
#print tree1,tree2
path1 = list(tree.leaf_treeposition(index1))
path2 = list(tree.leaf_treeposition(index2))
if index1 < index2:
dominate_path = tree.treeposition_spanning_leaves(index1,index2+1)
else:
dominate_path = tree.treeposition_spanning_leaves(index2,index1+1)
dept = len(dominate_path)
subtree = copy.deepcopy(tree)
#print index1,index2
#print dominate_path
for i in list(dominate_path):
temp = subtree[i]
subtree = temp
#print subtree
if subtree.node == 'NP':
if subtree[path1[dept]].node == 'NP' and subtree[path2[dept]].node == 'NP':
if path1[dept] == (path2[dept] - 2) and subtree[path1[dept]+1].node == ',':
return True
if path1[dept] == (path2[dept] + 2) and subtree[path1[dept]-1].node == ',':
return True
return False
def tree_distance(pair,parsed_sentences):
#within the same tree
if pair.first.sentenceID == pair.second.sentenceID:
index1 = eval(pair.first.offsets[0])
index2 = eval(pair.second.offsets[0])
#print index1,index2
#print pair.first.str,pair.second.str
tree = parsed_sentences[eval(pair.first.sentenceID)]
#print tree1,tree2
path1 = list(tree.leaf_treeposition(index1))
path2 = list(tree.leaf_treeposition(index2))
n = 0
for i,j in zip(path1,path2):
if i == j:
n+=1
else:
break
tree_distance = len(path1)+len(path2)-2*n
return tree_distance
else:
return "unknown"
def padding_leaves(tree):
leaves_location = [
tree.leaf_treeposition(i) for i in range(len(tree.leaves()))
]
for i in range(len(leaves_location)):
tree[leaves_location[i]] = "{0:03}".format(i) + "||||" + tree[
leaves_location[i]]
for i in range(len(tree.leaves())):
if len(tree[tree.leaf_treeposition(i)[:-1]]) > 1:
tree[tree.leaf_treeposition(i)] = Tree(
tree[tree.leaf_treeposition(i)[:-1]].label(),
[tree.leaves()[i]])
def padding_leaves_wnum(leaves, tree):
# leaves_location = [tree.leaf_treeposition(i) for i in range(len(tree.leaves()))]
# for i in range(len(leaves_location)):
# # tree[leaves_location[i]] = "{0:03}".format(i) + "||||" + tree[leaves_location[i]]
# tree[leaves_location[i]] = f'{i}'
# for i in range(len(tree.leaves())):
# if len(tree[tree.leaf_treeposition(i)[:-1]]) > 1:
# tree[tree.leaf_treeposition(i)] = Tree(tree[tree.leaf_treeposition(i)[:-1]].label(), [tree.leaves()[i]])
for i in range(len(leaves)):
tree[tree.leaf_treeposition(i)] = f'{i}'
def depth(entity,parsed_sentences):
index = eval(entity.offsets[0])
#print index1,index2
#print pair.first.str,pair.second.str
tree = parsed_sentences[eval(entity.sentenceID)]
#print tree1,tree2
path = list(tree.leaf_treeposition(index))
return len(path)
def get_PTP(pair,parsed_sentences):
m1_index = pair.first.offsets[0]
m2_index = pair.second.offsets[0]
senID = pair.first.sentenceID
tree = parsed_sentences[senID]
if m2_index >= len(tree.leaves()):
m2_index -=1
path1 = list(tree.leaf_treeposition(m1_index))
path2 = list(tree.leaf_treeposition(m2_index))
phrase_labels = []
n = 0
share_path = []
for i,j in zip(path1,path2):
if i == j:
n+=1
share_path.append(i)
else:
break
sub_path1 = path1[n:]
sub_path2 = path2[n:]
def get_labels(stree,path):
subtree = copy.deepcopy(stree)
labels = [subtree.node]
for i in path:
if isinstance(subtree[i],nltk.tree.Tree):
labels.append(subtree[i].node)
temp = subtree[i]
subtree = temp
return tuple((subtree,labels))
subtree = get_labels(tree,share_path)[0]
path1_labels = get_labels(subtree,sub_path1)[1]
path2_labels = get_labels(subtree,sub_path2)[1]
path1_labels.reverse()
path2_labels.reverse()
if path1_labels[-1] == path2_labels[0]:
return list(set(path1_labels[:]+path2_labels[1:]))
else:
ValueError("Path cannot connect:%s,%s" % (path1_labels,path2_labels ))
def remap_chars(tree):
for i in range(len(tree.leaves())):
if tree.leaves()[i] in SPECIAL_CHAR_MBACK:
tree[tree.leaf_treeposition(i)] = SPECIAL_CHAR_MBACK[tree.leaves()
[i]]