def initialization_feature(instance, use_gold=False): instance['relation'] = Relation.from_json(instance['relation_json']) relation = instance['relation'] instance['arg1_parse'] = json.loads(instance['arg1_parse_json']) instance['arg2_parse'] = json.loads(instance['arg2_parse_json']) if instance['arg1_parse']['dependency_tree'] is not None: instance['arg1_dependency_tree'] = DependencyGraph.from_dict(instance['arg1_parse']['dependency_tree']) else: instance['arg1_dependency_tree'] = None if instance['arg2_parse']['dependency_tree'] is not None: instance['arg2_dependency_tree'] = DependencyGraph.from_dict(instance['arg2_parse']['dependency_tree']) else: instance['arg2_dependency_tree'] = None instance['parse_tree1'] = Tree(instance['arg1_parse']['parsetree']) instance['parse_tree2'] = Tree(instance['arg2_parse']['parsetree']) instance['arg1_token_indices'] = \ get_indices_from_span_ranges(relation.arg1.text_span_ranges, instance['arg1_parse']) instance['arg2_token_indices'] = \ get_indices_from_span_ranges(relation.arg2.text_span_ranges, instance['arg2_parse']) instance['sentence1_tokens'] = [x[0] for x in instance['arg1_parse']['words']] instance['sentence2_tokens'] = [x[0] for x in instance['arg2_parse']['words']] instance['arg1_majority_head_verb_index'] = \ _get_majority_head_verb_index(instance['arg1_parse'], instance['arg1_token_indices']) instance['arg2_majority_head_verb_index'] = \ _get_majority_head_verb_index(instance['arg2_parse'], instance['arg2_token_indices']) return []
def convert_dependencies_to_graph(dependencies):
graph = DependencyGraph()
root_node = None
assert (len(dependencies) > 0)
#Try to find root
for dep in dependencies:
if dep[0] == 'root':
root_node = Node(dep[1])
first_real_node = Node(dep[2])
root_edge = Edge(root_node, first_real_node, dep[0])
root_node.edges.append(root_edge)
graph.root_node = root_node
graph.add_node(root_node)
graph.add_node(first_real_node)
graph.add_edge(root_edge)
break
if root_node is not None:
_construct_subgraph(dependencies, first_real_node, graph)
return graph
def convert_dependencies_to_graph(dependencies): graph = DependencyGraph() root_node = None assert(len(dependencies) > 0) #Try to find root for dep in dependencies: if dep[0] == 'root': root_node = Node(dep[1]) first_real_node = Node(dep[2]) root_edge = Edge(root_node, first_real_node, dep[0]) root_node.edges.append(root_edge) graph.root_node = root_node graph.add_node(root_node) graph.add_node(first_real_node) graph.add_edge(root_edge) break if root_node is not None: _construct_subgraph(dependencies, first_real_node, graph) return graph
return graph
def _construct_subgraph(dependencies, root_node, graph):
for dep in dependencies:
if dep[1] == root_node.token:
new_node = Node(dep[2])
new_edge = Edge(root_node, new_node, dep[0])
root_node.edges.append(new_edge)
graph.add_node(new_node)
graph.add_edge(new_edge)
_construct_subgraph(dependencies, new_node, graph)
import json
from tpl.misc import TJsonEncoder
if __name__ == '__main__':
"""For testing purposes"""
from tpl.language.stanford_parser import Parser
parser = Parser()
parse_result = parser.parse(
'Barack Obama, who you talked to yesterday, play basketball.')
sentence = parse_result['sentences'][0]
split_clauses_from_sentence(sentence)
print json.dumps(sentence, indent=2, cls=TJsonEncoder)
print json.dumps(sentence['dependency_tree'].to_dict(), indent=2)
dgraph = DependencyGraph.from_dict(sentence['dependency_tree'].to_dict())
print json.dumps(dgraph.to_dict(), indent=2)
print dgraph.to_dict() == DependencyGraph.from_dict(
dgraph.to_dict()).to_dict()
graph.add_edge(root_edge)
break
if root_node is not None:
_construct_subgraph(dependencies, first_real_node, graph)
return graph
def _construct_subgraph(dependencies, root_node, graph):
for dep in dependencies:
if dep[1] == root_node.token:
new_node = Node(dep[2])
new_edge = Edge(root_node, new_node, dep[0])
root_node.edges.append(new_edge)
graph.add_node(new_node)
graph.add_edge(new_edge)
_construct_subgraph(dependencies, new_node, graph)
import json
from tpl.misc import TJsonEncoder
if __name__ == '__main__':
"""For testing purposes"""
from tpl.language.stanford_parser import Parser
parser = Parser()
parse_result = parser.parse('Barack Obama, who you talked to yesterday, play basketball.')
sentence = parse_result['sentences'][0]
split_clauses_from_sentence(sentence)
print json.dumps(sentence, indent=2, cls=TJsonEncoder)
print json.dumps(sentence['dependency_tree'].to_dict(), indent=2)
dgraph = DependencyGraph.from_dict(sentence['dependency_tree'].to_dict())
print json.dumps(dgraph.to_dict(), indent=2)
print dgraph.to_dict() == DependencyGraph.from_dict(dgraph.to_dict()).to_dict()
