def get_constituents_with_label(parse_dict, connective):
DocID = connective.DocID
sent_index = connective.sent_index
parse_tree = parse_dict[DocID]["sentences"][sent_index]["parsetree"].strip(
)
syntax_tree = Syntax_tree(parse_tree)
if syntax_tree.tree == None:
return []
conn_indices = connective.token_indices
constituent_nodes = []
if len(conn_indices) == 1: # like and or so...
conn_node = syntax_tree.get_leaf_node_by_token_index(
conn_indices[0]).up
else:
conn_node = syntax_tree.get_common_ancestor_by_token_indices(
conn_indices)
conn_leaves = set([
syntax_tree.get_leaf_node_by_token_index(conn_index)
for conn_index in conn_indices
])
children = conn_node.get_children()
for child in children:
leaves = set(child.get_leaves())
if conn_leaves & leaves == set([]):
constituent_nodes.append(child)
curr = conn_node
while not curr.is_root():
constituent_nodes.extend(syntax_tree.get_siblings(curr))
curr = curr.up
Arg1_token_indices = connective.Arg1_token_indices
Arg2_token_indices = connective.Arg2_token_indices
Arg1_leaves = set([
syntax_tree.get_leaf_node_by_token_index(index)
for index in Arg1_token_indices
])
Arg2_leaves = set([
syntax_tree.get_leaf_node_by_token_index(index)
for index in Arg2_token_indices
])
# 根据node生成Constituent对象，并标记
constituents = []
for node in constituent_nodes:
cons = Constituent(syntax_tree, node)
cons.connective = connective
leaves = set(node.get_leaves())
if leaves <= Arg1_leaves:
cons.label = "Arg1"
elif leaves <= Arg2_leaves:
cons.label = "Arg2"
else:
cons.label = "NULL"
constituents.append(cons)
return constituents
def _get_constituents(parse_dict, connective):
DocID = connective.DocID
sent_index = connective.sent_index
parse_tree = parse_dict[DocID]["sentences"][sent_index]["parsetree"].strip()
syntax_tree = Syntax_tree(parse_tree)
if syntax_tree.tree == None:
return []
conn_indices = connective.token_indices
constituent_nodes = []
if len(conn_indices) == 1:# like and or so...
conn_node = syntax_tree.get_leaf_node_by_token_index(conn_indices[0]).up
else:
conn_node = syntax_tree.get_common_ancestor_by_token_indices(conn_indices)
conn_leaves = set([syntax_tree.get_leaf_node_by_token_index(conn_index) for conn_index in conn_indices])
children = conn_node.get_children()
for child in children:
leaves = set(child.get_leaves())
if conn_leaves & leaves == set([]):
constituent_nodes.append(child)
curr = conn_node
while not curr.is_root():
constituent_nodes.extend(syntax_tree.get_siblings(curr))
curr = curr.up
# obtain the Constituent object according to the node.
constituents = []
for node in constituent_nodes:
cons = Constituent(syntax_tree, node)
cons.connective = connective
constituents.append(cons)
return constituents
def get_constituents_with_label2(parse_dict, connective):
DocID = connective.DocID
sent_index = connective.sent_index
parse_tree = parse_dict[DocID]["sentences"][sent_index]["parsetree"].strip()
syntax_tree = Syntax_tree(parse_tree)
if syntax_tree.tree == None:
return []
conn_indices = connective.token_indices
constituent_nodes = []
if len(conn_indices) == 1:# like and or so...
conn_node = syntax_tree.get_leaf_node_by_token_index(conn_indices[0]).up
else:
conn_node = syntax_tree.get_common_ancestor_by_token_indices(conn_indices)
conn_leaves = set([syntax_tree.get_leaf_node_by_token_index(conn_index) for conn_index in conn_indices])
children = conn_node.get_children()
for child in children:
leaves = set(child.get_leaves())
if conn_leaves & leaves == set([]):
constituent_nodes.append(child)
curr = conn_node
while not curr.is_root():
constituent_nodes.extend(syntax_tree.get_siblings(curr))
curr = curr.up
Arg1_token_indices = connective.Arg1_token_indices
Arg2_token_indices = connective.Arg2_token_indices
Arg1_leaves = set([syntax_tree.get_leaf_node_by_token_index(index) for index in Arg1_token_indices])
Arg2_leaves = set([syntax_tree.get_leaf_node_by_token_index(index) for index in Arg2_token_indices])
constituents = []
for node in constituent_nodes:
cons = Constituent(syntax_tree, node)
cons.connective = connective
leaves = set(node.get_leaves())
if leaves <= Arg1_leaves:
cons.label = "Arg1"
elif leaves <= Arg2_leaves:
cons.label = "Arg2"
else:
cons.label = "NULL"
constituents.append(cons)
return constituents
def ssArgumentExt(inputFilenamePath):
parse_file = codecs.open(inputFilenamePath+'/parses.json', encoding='utf8');
en_parse_dict = json.load(parse_file);
i = 0;
for prediction in observedArray:
filename = bigDiction[i][2];
sentenceNumber = int(bigDiction[i+1][3]) + 1;
connWordID = int(bigDiction[i][4]);
print "ConnWordID: " + str(connWordID);
parse_tree = en_parse_dict[filename]["sentences"][sentenceNumber]["parsetree"].strip();
syntax_tree = Syntax_tree(parse_tree)
if syntax_tree.tree == None:
return []
#Get Connective Indices
conn_indices = [connWordID];
constituent_nodes = [];
if len(conn_indices) == 1:# like and or so...
conn_node = syntax_tree.get_leaf_node_by_token_index(conn_indices[0]).up
else:
conn_node = syntax_tree.get_common_ancestor_by_token_indices(conn_indices)
conn_leaves = set([syntax_tree.get_leaf_node_by_token_index(conn_index) for conn_index in conn_indices])
children = conn_node.get_children()
for child in children:
leaves = set(child.get_leaves())
if conn_leaves & leaves == set([]):
constituent_nodes.append(child)
curr = conn_node
while not curr.is_root():
constituent_nodes.extend(syntax_tree.get_siblings(curr))
curr = curr.up
# obtain the Constituent object according to the node.
constituents = []
for node in constituent_nodes:
cons = Constituent(syntax_tree, node)
#print "Object Type: " + str(cons.type());
#print "Object Dir: " + str(cons.dir());
#print "Object id: " + str(cons.id());
#print "cons: " + str(cons.connective);
connective = Connective(filename, sentenceNumber, conn_indices, "text");
cons.connective = connective
constituents.append(cons)
i = i + 1;
print "Connective ID:" + str(connWordID);
print "Size of Observed Array: " + str(len(observedArray));
print "Size of Constituents Array: " + str(len(constituents));
