def filter_second_hops(grounded_graphs):
"""
This methods filters out second hop relations that are already present as first hop relations. Relation direction is
respected.
:param grounded_graphs: list of grounded graphs
:return: filtered list of grounded graphs
>>> filter_second_hops([ {'edgeSet': [{'kbID': 'P17v','type': 'direct'}]}, {'edgeSet': [{'kbID': 'P17v', 'hopUp':'P31v', 'type': 'direct'}]} ])
[{'edgeSet': [{'kbID': 'P17v', 'type': 'direct'}]}]
>>> filter_second_hops([ {'edgeSet': [{'kbID': 'P17v','type': 'direct'}]}, {'edgeSet': [{'type':'time', 'argmax':'time'}]} ])
[{'edgeSet': [{'kbID': 'P17v', 'type': 'direct'}]}, {'edgeSet': [{'type': 'time', 'argmax': 'time'}]}]
>>> filter_second_hops([ {'edgeSet': [{'kbID': 'P17v','type': 'direct'}]}, {'edgeSet': [{'kbID': 'P17v','type': 'direct'}]} ])
[{'edgeSet': [{'kbID': 'P17v', 'type': 'direct'}]}, {'edgeSet': [{'kbID': 'P17v', 'type': 'direct'}]}]
"""
first_order_relations = {
"{}-{}".format(
graph.get_graph_last_edge(g).get('kbID'),
graph.get_graph_last_edge(g).get('type'))
for g in grounded_graphs if graph.get_graph_last_edge(g).get(
'type') in {'direct', 'reverse', 'v-structure'}
and 'hopUp' not in graph.get_graph_last_edge(g)
and 'hopDown' not in graph.get_graph_last_edge(g)
}
grounded_graphs = [
g for g in grounded_graphs
if ('hopUp' not in graph.get_graph_last_edge(g) and 'hopDown' not in
graph.get_graph_last_edge(g)) or ("{}-{}".format(
graph.get_graph_last_edge(g).get('kbID'),
graph.get_graph_last_edge(g).get('type')) not in
first_order_relations)
]
logger.debug("Filter out unnecessary hops: {}".format(
len(grounded_graphs)))
return grounded_graphs
def filter_qualifier_relations(grounded_graphs):
"""
Filter out relation so that we always prefer the direct relation over qualifiers. Relation direction is not
respected.
:param grounded_graphs: list of grounded graphs
:return: filtered list of grounded graphs
>>> filter_qualifier_relations([ {'edgeSet': [{'kbID': 'P17v','type': 'direct'}]}, {'edgeSet': [{'kbID': 'P17q', 'type': 'direct'}]} ])
[{'edgeSet': [{'kbID': 'P17v', 'type': 'direct'}]}]
>>> filter_qualifier_relations([ {'edgeSet': [{'kbID': 'P17v','type': 'direct'}]}, {'edgeSet': [{'kbID': 'P17q', 'type': 'reverse'}]} ])
[{'edgeSet': [{'kbID': 'P17v', 'type': 'direct'}]}]
>>> filter_qualifier_relations([ {'edgeSet': [{'kbID': 'P17v','type': 'direct'}]}, {'edgeSet': [{'kbID': 'P17q', 'hopUp':'P31v', 'type': 'direct'}]} ])
[{'edgeSet': [{'kbID': 'P17v', 'type': 'direct'}]}]
>>> filter_qualifier_relations([ {'edgeSet': [{'kbID': 'P175q','type': 'v-structure'}]}, {'edgeSet': [{'type':'direct', 'kbID':'P175q'}]} ])
[{'edgeSet': [{'type': 'v-structure', 'kbID': 'P175q'}]}]
>>> filter_qualifier_relations([ {'edgeSet': [{'kbID': 'P17v','type': 'direct'}]}, {'edgeSet': [{'type':'time', 'argmax':'time'}]} ])
[{'edgeSet': [{'kbID': 'P17v', 'type': 'direct'}]}, {'edgeSet': [{'type': 'time', 'argmax': 'time'}]}]
"""
direct_relations = {
graph.get_graph_last_edge(g).get('kbID', "")[:-1]
for g in grounded_graphs
if (graph.get_graph_last_edge(g).get('type') in {'direct', 'reverse'}
and graph.get_graph_last_edge(g).get('kbID', " ")[-1] not in "qr")
or graph.get_graph_last_edge(g).get('type') == 'v-structure'
}
grounded_graphs = [
g for g in grounded_graphs if graph.get_graph_last_edge(g).get('type')
not in {'direct', 'reverse'} or graph.get_graph_last_edge(g).get(
'kbID', " ")[-1] not in "qr" or graph.get_graph_last_edge(g).get(
'kbID', "")[:-1] not in direct_relations
]
logger.debug("Filter out unnecessary qualifiers: {}".format(
len(grounded_graphs)))
return grounded_graphs
def filter_repeated_relations(grounded_graphs):
"""
:param grounded_graphs: list of grounded graphs
:return: filtered list of grounded graphs
>>> filter_repeated_relations([ {'edgeSet': [{'kbID': 'P17v','type': 'direct', 'rightkbID':'Q567'}]}, {'edgeSet': [{'kbID': 'P17v','type': 'direct', 'rightkbID':'Q1234'}]} ])
[{'edgeSet': [{'kbID': 'P17v', 'type': 'direct', 'rightkbID': 'Q567'}]}]
"""
repeated_relations = set()
filtered_graphs = []
for g in grounded_graphs:
edge = graph.get_graph_last_edge(g)
edge_str = "{}-{}".format(edge.get('kbID'), edge.get('type'))
if not edge_str in repeated_relations:
filtered_graphs.append(g)
repeated_relations.add(edge_str)
logger.debug("Filter out repeated relations: {}".format(
len(filtered_graphs)))
return filtered_graphs
def last_relation_numeric(g):
"""
Adds a numeric restriction to the last relation in the graph.
:param g: a graph with a non-empty edgeSet
:return: a list of suggested graphs
>>> last_relation_numeric({'edgeSet': [{'right':[2], 'type':'direct'}, {'right':[8], 'type':'direct'}], 'entities': [{'linkings':[("Q37876", "Natalie Portman")], 'tokens':["Portman"], 'type':'PERSON'}, {'linkings': [(None, ['2012'])], 'type': 'CD', 'tokens': ['2012']}]}) == \
[{'edgeSet': [{'right':[2], 'type':'direct'}, {'right':[8], 'type':'direct', 'num': ['2012']}], 'entities': [{'linkings':[("Q37876", "Natalie Portman")], 'tokens':["Portman"], 'type':'PERSON'}]}]
True
>>> last_relation_numeric({'edgeSet': [{'right':[2]}, {'right':[8], 'argmin':'time'}], 'entities': [{'linkings': [(None, ['2012'])], 'type': 'CD', 'tokens': ['2012']}]})
[]
>>> last_relation_numeric({'edgeSet': [{'right':[2]}, {'right':[8], 'num':'2009'}], 'entities': [{'linkings': [(None, ['2012'])], 'type': 'CD', 'tokens': ['2012']}]})
[]
>>> last_relation_numeric({'edgeSet': [{'right':[2]}], 'entities': []})
[]
>>> last_relation_numeric({'edgeSet': [{'right':[2]}], 'entities': [{'linkings':[("Q37876", "Natalie Portman")], 'tokens':["Portman"], 'type':'PERSON'}]})
[]
"""
if len(g.get('edgeSet', [])) == 0 or graph.graph_has_temporal(g):
return []
if len(g.get('entities', [])) == 0 or not any(
e.get("type") == 'CD' for e in g['entities'] if len(e) > 1):
return []
entities = copy.copy(g.get('entities', []))
cd_entities = [e.get("tokens") for e in entities if e.get("type") == 'CD']
if len(cd_entities) == 0:
return []
cd_entity = cd_entities[0]
entities = [e for e in entities if e.get("tokens") != cd_entity]
new_g = graph.copy_graph(g)
new_g['entities'] = entities
edge_to_modify = graph.get_graph_last_edge(
new_g, filter_out_types={'iclass', 'v-structure', 'class'})
if len(edge_to_modify) > 0:
edge_to_modify['num'] = cd_entity
return [new_g]
return []
def last_relation_hop(g):
"""
Takes a graph with an existing relation and an intermediate variable by performing a hop-up for the second entity.
:param g: a graph with an non-empty edgeSet
:return: a list of suggested graphs
>>> last_relation_hop({'edgeSet': [], 'entities': [[4, 5, 6]]})
[]
>>> last_relation_hop({'edgeSet': [{'right':[4,5,6]}], 'entities': []}) ==\
[{'edgeSet': [{'right':[4,5,6], 'hopUp': None}], 'entities': []}, {'edgeSet': [{'right':[4,5,6], 'hopDown': None}], 'entities': []}]
True
>>> last_relation_hop({'edgeSet': [{'right':[4,5,6], 'hopUp': None}], 'entities': []})
[]
>>> last_relation_hop({'edgeSet': [{'right':[4,5,6], 'kbID': "P31v", "type": "direct"}], 'entities': []})
[]
>>> last_relation_hop({'edgeSet': [{'right':["Bahama"], "rightkbID":"Q6754", 'type':'direct'}], 'entities': []}) ==\
[{'edgeSet': [{'right':["Bahama"], "rightkbID":"Q6754", 'hopUp': None, 'type':'direct'}], 'entities': []}, {'edgeSet': [{'right':["Bahama"], "rightkbID":"Q6754", 'hopDown': None, 'type':'direct'}], 'entities': []}]
True
>>> last_relation_hop({'edgeSet': [{'right':[4,5,6], 'argmax':'time'}], 'entities': []})
[]
>>> last_relation_hop({'edgeSet': [{'right':[4,5,6], 'num':['2012']}], 'entities': []})
[]
"""
if len(g.get('edgeSet', [])) == 0 or any(
hop in g['edgeSet'][-1]
for hop in {'hopUp', 'hopDown'}) or graph.graph_has_temporal(g):
return []
new_graphs = []
for hop in HOP_TYPES:
new_g = graph.copy_graph(g)
edge_to_modify = graph.get_graph_last_edge(
new_g, filter_out_types={'iclass', 'class', 'v-structure', 'time'})
if len(edge_to_modify) > 0 and 'kbID' not in edge_to_modify:
edge_to_modify[hop] = None
new_graphs.append(new_g)
return new_graphs