def test_has_get():
g = _read_graph()
assert semgraph.has_out_edge(g, 'T0', {"punct"})
assert not semgraph.has_out_edge(g, 'T2', {"punct"})
assert not semgraph.has_out_edge(g, 'T0', {""})
assert semgraph.has_in_edge(g, 'T1', {"punct"})
assert not semgraph.has_in_edge(g, 'T2', {"punct"})
assert not semgraph.has_in_edge(g, 'T1', {""})
assert semgraph.has_out_node(g, 'T0', {":"})
assert not semgraph.has_out_node(g, 'T1', {":"})
assert not semgraph.has_out_node(g, 'T0', {""})
assert semgraph.has_in_node(g, 'T1', {"finding"})
assert not semgraph.has_in_node(g, 'T4', {"finding"})
assert not semgraph.has_in_node(g, 'T1', {""})
assert semgraph.get_in(g, 'T1', {"finding"}, {"punct"})
assert semgraph.get_in(g, 'T1', {""}, {"punct"}) is None
assert semgraph.has_in(g, 'T1', {"finding"}, {"punct"})
assert not semgraph.has_in(g, 'T1', {""}, {"punct"})
assert semgraph.has_out(g, 'T0', {":"}, {"punct"})
assert not semgraph.has_out(g, 'T0', {""}, {"punct"})
def match_uncertainty(self, graph, node):
for pattern in self.uncertain_patterns:
for m in pattern.finditer(graph):
n0 = m.group(0)
if n0 == node:
return m
# parsing error
# suggestive of XXX
p = ngrex.compile('{} <{dependency:/nmod:of/} {lemma:/suggestive/}')
for m in p.finditer(graph):
n0 = m.group(0)
if n0 == node:
if semgraph.has_out_node(graph, m.group(1), ['most']):
return None
elif semgraph.has_out(graph, n0, ['new', 'develop'], ['amod']):
continue
else:
return m
return None
def propagate(G):
for i in range(0, 2):
edges = []
for node in G.nodes():
# hypoinflated but clear of
if G.nodes[node]['lemma'] == 'hypoinflated':
for child in G.successors(node):
edge_dep = G[node][child]['dependency']
if G.nodes[child][
'lemma'] == 'clear' and edge_dep == 'conj:but':
for of in G.successors(node):
of_dep = G[node][of]['dependency']
if of_dep == 'nmod:of':
edges.append(Edge(child, of, of_dep))
break
for p, c, d in G.edges(data=True):
# propagate appos
if d['dependency'] == 'appos':
# x > y >appos > z
for grandpa in G.predecessors(p):
edge_dep = G[grandpa][p]['dependency']
edges.append(Edge(grandpa, c, edge_dep))
# x <neg < y >appos > z
for child in G.successors(p):
edge_dep = G[p][child]['dependency']
if edge_dep == 'neg':
edges.append(Edge(c, child, edge_dep))
# propagate dep
if d['dependency'] == 'dep' \
and G.nodes[p]['tag'].startswith('N') \
and G.nodes[c]['tag'].startswith('N'):
for grandchild in G.successors(c):
edge_dep = G[c][grandchild]['dependency']
if edge_dep == 'neg':
edges.append(Edge(p, grandchild, edge_dep))
# propagate cop conjunction
if d['dependency'].startswith('conj') \
and G.nodes[p]['tag'].startswith('N') \
and G.nodes[c]['tag'].startswith('N'):
for child in G.successors(p):
edge_dep = G[p][child]['dependency']
if edge_dep in ('aux', 'cop', 'neg', 'amod'):
edges.append(Edge(c, child, edge_dep))
if edge_dep in ('dep', 'compound'
) and G.nodes[child]['lemma'] == 'no':
edges.append(Edge(c, child, edge_dep))
if edge_dep == 'case' and G.nodes[child][
'lemma'] == 'without':
edges.append(Edge(c, child, edge_dep))
# propagate area/amount >of XXX
if d['dependency'] == 'nmod:of' and G.nodes[p]['lemma'] in (
'area', 'amount'):
for grandpa in G.predecessors(p):
edge_dep = G[grandpa][p]['dependency']
edges.append(Edge(grandpa, c, edge_dep))
# propagate combination of XXX
if d['dependency'] == 'nmod:of' and G.nodes[p][
'lemma'] == 'combination':
for grandpa in G.predecessors(p):
edge_dep = G[grandpa][p]['dependency']
edges.append(Edge(grandpa, c, edge_dep))
if d['dependency'] == 'nmod:of':
for child in G.successors(p):
edge_dep = G[p][child]['dependency']
# propagate no <neg x >of XXX
if edge_dep == 'neg':
edges.append(Edge(c, child, edge_dep))
# propagate without <case x >of XXX
if edge_dep == 'case' and G.nodes[child] == 'without':
edges.append(Edge(c, child, edge_dep))
# parse error
# no xx and xxx
if d['dependency'] == 'neg' and semgraph.has_out_node(
G, p, ['or', 'and']):
for child in G.successors(p):
edge_dep = G[p][child]['dependency']
if edge_dep == 'compound' and G.nodes[child][
'tag'].startswith('N'):
edges.append(Edge(child, c, 'neg'))
has_more_edges = False
for e in edges:
if not G.has_edge(e.gov, e.dep):
assert isinstance(e.data, str) or isinstance(
e.data, unicode), type(e.data)
G.add_edge(e.gov, e.dep, dependency=e.data)
has_more_edges = True
if not has_more_edges:
break
def propagate(G):
logger = logging.getLogger(__name__)
for i in range(0, 2):
edges = []
for node in G.nodes_iter():
# hypoinflated but clear of
if G.node[node]['lemma'] == 'hypoinflated':
for child in G.successors_iter(node):
edge_dep = G.edge[node][child]['dependency']
if G.node[child][
'lemma'] == 'clear' and edge_dep == 'conj:but':
for of in G.successors_iter(node):
of_dep = G.edge[node][of]['dependency']
if of_dep == 'nmod:of':
edges.append((child, of, {
'dependency': of_dep
}))
break
for p, c, d in G.edges_iter(data=True):
# propagate appos
if d['dependency'] == 'appos':
# x > y >appos > z
for grandpa in G.predecessors_iter(p):
edge_dep = G[grandpa][p]['dependency']
edges.append((grandpa, c, {'dependency': edge_dep}))
# x <neg < y >appos > z
for child in G.successors_iter(p):
edge_dep = G.edge[p][child]['dependency']
if edge_dep == 'neg':
edges.append((c, child, {'dependency': edge_dep}))
# propagate dep
if d['dependency'] == 'dep' \
and G.node[p]['tag'].startswith('N') \
and G.node[c]['tag'].startswith('N'):
for grandchild in G.successors_iter(c):
edge_dep = G.edge[c][grandchild]['dependency']
if edge_dep == 'neg':
edges.append((p, grandchild, {'dependency': edge_dep}))
# propagate cop conjunction
if d['dependency'].startswith('conj') \
and G.node[p]['tag'].startswith('N') \
and G.node[c]['tag'].startswith('N'):
for child in G.successors_iter(p):
edge_dep = G.edge[p][child]['dependency']
if edge_dep in ('aux', 'cop', 'neg', 'amod'):
edges.append((c, child, {'dependency': edge_dep}))
if edge_dep in ('dep', 'compound'
) and G.node[child]['lemma'] == 'no':
edges.append((c, child, {'dependency': edge_dep}))
if edge_dep == 'case' and G.node[child][
'lemma'] == 'without':
edges.append((c, child, {'dependency': edge_dep}))
# propagate area/amount >of XXX
if d['dependency'] == 'nmod:of' and G.node[p]['lemma'] in (
'area', 'amount'):
for grandpa in G.predecessors_iter(p):
edge_dep = G[grandpa][p]['dependency']
edges.append((grandpa, c, {'dependency': edge_dep}))
# propagate combination of XXX
if d['dependency'] == 'nmod:of' and G.node[p][
'lemma'] == 'combination':
for grandpa in G.predecessors_iter(p):
edge_dep = G[grandpa][p]['dependency']
edges.append((grandpa, c, {'dependency': edge_dep}))
if d['dependency'] == 'nmod:of':
for child in G.successors_iter(p):
edge_dep = G.edge[p][child]['dependency']
# propagate no <neg x >of XXX
if edge_dep == 'neg':
edges.append((c, child, {'dependency': edge_dep}))
# propagate without <case x >of XXX
if edge_dep == 'case' and G.node[child] == 'without':
edges.append((c, child, {'dependency': edge_dep}))
# parse error
# no xx and xxx
if d['dependency'] == 'neg' and semgraph.has_out_node(
G, p, ['or', 'and']):
for child in G.successors_iter(p):
edge_dep = G.edge[p][child]['dependency']
if edge_dep == 'compound' and G.node[child][
'tag'].startswith('N'):
edges.append((child, c, {'dependency': 'neg'}))
has_more_edges = False
for p, c, d in edges:
if not G.has_edge(p, c):
G.add_edge(p, c, attr_dict=d)
has_more_edges = True
if not has_more_edges:
break