def _make_context(self, entity):
"""Get place and time info from the json for this entity."""
loc_context = None
time_context = None
# Look for time and place contexts.
for argument in entity["arguments"]:
if argument["type"] == "place":
entity_id = argument["value"]["@id"]
loc_entity = self.concept_dict[entity_id]
place = loc_entity["canonicalName"]
geo_id = loc_entity.get('geoname_id')
loc_context = RefContext(name=place, db_refs={"GEOID": geo_id})
if argument["type"] == "time":
entity_id = argument["value"]["@id"]
temporal_entity = self.concept_dict[entity_id]
text = temporal_entity['mentions'][0]['text']
if len(temporal_entity.get("timeInterval", [])) < 1:
time_context = TimeContext(text=text)
continue
time = temporal_entity["timeInterval"][0]
start = datetime.strptime(time['start'], '%Y-%m-%dT%H:%M')
end = datetime.strptime(time['end'], '%Y-%m-%dT%H:%M')
duration = int(time['duration'])
time_context = TimeContext(text=text,
start=start,
end=end,
duration=duration)
# Put context together
context = None
if loc_context or time_context:
context = WorldContext(time=time_context, geo_location=loc_context)
return context
def _resolve_time(hume_temporal_entity):
text = hume_temporal_entity['mentions'][0]['text']
if len(hume_temporal_entity.get("timeInterval", [])) < 1:
return TimeContext(text=text)
time = hume_temporal_entity["timeInterval"][0]
start = datetime.strptime(time['start'], '%Y-%m-%dT%H:%M')
end = datetime.strptime(time['end'], '%Y-%m-%dT%H:%M')
duration = int(time['duration'])
return TimeContext(text=text, start=start, end=end, duration=duration)
def _resolve_time(hume_temporal_entity):
text = hume_temporal_entity['mentions'][0]['text']
if len(hume_temporal_entity.get("timeInterval", [])) < 1:
return TimeContext(text=text)
time = hume_temporal_entity["timeInterval"][0]
start = datetime.strptime(time['start'], '%Y-%m-%dT%H:%M')
end = datetime.strptime(time['end'], '%Y-%m-%dT%H:%M')
end = end + timedelta(minutes=1)
duration = int((end - start).total_seconds())
return TimeContext(text=text, start=start, end=end, duration=duration)
def get_event(event_entry):
name = event_entry['Relation']
concept = Concept(name, db_refs={'TEXT': name})
grounding = event_entry['Event_Type']
if grounding:
concept.db_refs['SOFIA'] = grounding
context = WorldContext()
time = event_entry.get('Time')
if time:
context.time = TimeContext(text=time.strip())
loc = event_entry.get('Location')
if loc:
context.geo_location = RefContext(name=loc)
text = event_entry.get('Text')
ref = event_entry.get('Source')
agent = event_entry.get('Agent')
patient = event_entry.get('Patient')
anns = {}
if agent:
anns['agent'] = agent
if patient:
anns['patient'] = patient
ev = Evidence(source_api='sofia', pmid=ref, text=text,
annotations=anns, source_id=event_entry['Event Index'])
pol = event_entry.get('Polarity')
event = Event(concept, context=context, evidence=[ev],
delta=QualitativeDelta(polarity=pol, adjectives=None))
return event
def get_event_compositional(self, event_entry: Dict[str, str]) -> Event:
"""Get an Event with compositional grounding
Parameters
----------
event_entry :
The event to process
Returns
-------
event :
An Event statement
"""
# Get get compositional grounding
comp_name, comp_grnd = self.get_compositional_grounding(event_entry)
if comp_name is not None and \
comp_grnd[0] is not None and \
comp_grnd[0][0] is not None:
concept = Concept(comp_name,
db_refs={
'TEXT': comp_name,
'WM': [comp_grnd]
})
# If not try to get old style Sofia grounding
else:
name = event_entry['Relation']
concept = Concept(name, db_refs={'TEXT': name})
if event_entry['Event_Type']:
concept.db_refs['SOFIA'] = event_entry['Event_Type']
context = WorldContext()
time = event_entry.get('Time')
if time:
context.time = TimeContext(text=time.strip())
loc = event_entry.get('Location')
if loc:
context.geo_location = RefContext(name=loc)
text = event_entry.get('Text')
ref = event_entry.get('Source')
agent = event_entry.get('Agent')
patient = event_entry.get('Patient')
anns = {}
if agent:
anns['agent'] = agent
if patient:
anns['patient'] = patient
text_refs = {'DART': ref}
ev = Evidence(source_api='sofia',
text_refs=text_refs,
text=text,
annotations=anns,
source_id=event_entry['Event Index'])
pol = event_entry.get('Polarity')
event = Event(concept,
context=context,
evidence=[ev],
delta=QualitativeDelta(polarity=pol, adjectives=None))
return event
def get_event(event_entry):
name = event_entry['Relation']
concept = Concept(name, db_refs={'TEXT': name})
grounding = event_entry['Event_Type']
if grounding:
concept.db_refs['SOFIA'] = grounding
context = WorldContext()
time = event_entry.get('Time')
if time:
context.time = TimeContext(text=time.strip())
loc = event_entry.get('Location')
if loc:
context.geo_location = RefContext(name=loc)
text = event_entry.get('Text')
ref = event_entry.get('Source')
ev = Evidence(source_api='sofia', pmid=ref, text=text)
pol = event_entry.get('Polarity')
event = Event(concept,
context=context,
evidence=[ev],
delta={
'polarity': pol,
'adjectives': []
})
return event
def time_context_from_dct(dct):
"""Return a time context object given a DCT entry."""
time_text = dct.get('text')
start = _get_time_stamp(dct.get('start'))
end = _get_time_stamp(dct.get('end'))
duration = dct.get('duration')
tc = TimeContext(text=time_text,
start=start,
end=end,
duration=duration)
return tc
def time_context_from_timex(timex):
"""Return a TimeContext object given a timex entry."""
time_text = timex.get('text')
constraint = timex['intervals'][0]
start = _get_time_stamp(constraint.get('start'))
end = _get_time_stamp(constraint.get('end'))
duration = constraint['duration']
tc = TimeContext(text=time_text,
start=start,
end=end,
duration=duration)
return tc
def time_context_from_timex(timex):
"""Return a TimeContext object given a timex entry."""
time_text = timex.get('text')
intervals = timex.get('intervals')
if not intervals:
start = end = duration = None
else:
constraint = intervals[0]
start = _get_time_stamp(constraint.get('start'))
end = _get_time_stamp(constraint.get('end'))
duration = _get_duration(start, end)
tc = TimeContext(text=time_text, start=start, end=end, duration=duration)
return tc
def get_event_flat(self, event_entry: Dict[str, str]) -> Event:
"""Get an Event with flattened grounding
Parameters
----------
event_entry :
The event to process
Returns
-------
event :
An Event statement
"""
name = event_entry['Relation']
concept = Concept(name, db_refs={'TEXT': name})
grounding = event_entry['Event_Type']
if grounding:
concept.db_refs['SOFIA'] = grounding
context = WorldContext()
time = event_entry.get('Time')
if time:
context.time = TimeContext(text=time.strip())
loc = event_entry.get('Location')
if loc:
context.geo_location = RefContext(name=loc)
text = event_entry.get('Text')
ref = event_entry.get('Source')
agent = event_entry.get('Agent')
patient = event_entry.get('Patient')
anns = {}
if agent:
anns['agent'] = agent
if patient:
anns['patient'] = patient
text_refs = {'DART': ref}
ev = Evidence(source_api='sofia',
text_refs=text_refs,
text=text,
annotations=anns,
source_id=event_entry['Event Index'])
pol = event_entry.get('Polarity')
event = Event(concept,
context=context,
evidence=[ev],
delta=QualitativeDelta(polarity=pol, adjectives=None))
return event
def _build_stmts(self, rel_dict):
stmt_list = []
cause_entries = rel_dict.get('Cause Index')
effect_entries = rel_dict.get('Effect Index')
# FIXME: Handle cases in which there is a missing cause/effect
if not cause_entries or not effect_entries:
return []
causes = [c.strip() for c in cause_entries.split(',')]
effects = [e.strip() for e in effect_entries.split(',')]
rel = rel_dict.get('Relation')
if _in_rels(rel, pos_rels):
pol = 1
elif _in_rels(rel, neg_rels):
pol = -1
elif _in_rels(rel, neu_rels):
pol = None
# If we don't recognize this relation, we don't get any
# statements
else:
return []
text = rel_dict.get('Sentence')
annot_keys = ['Relation']
annots = {k: rel_dict.get(k) for k in annot_keys}
ref = rel_dict.get('Source_File')
for cause_idx, effect_idx in itertools.product(causes, effects):
cause_name = self._events[cause_idx]['Relation']
cause_grounding = self._events[cause_idx]['Event_Type']
effect_name = self._events[effect_idx]['Relation']
effect_grounding = self._events[effect_idx]['Event_Type']
cause_concept = Concept(cause_name, db_refs={'TEXT': cause_name})
if cause_grounding:
cause_concept.db_refs['SOFIA'] = cause_grounding
effect_concept = Concept(effect_name,
db_refs={'TEXT': effect_name})
if effect_grounding:
effect_concept.db_refs['SOFIA'] = effect_grounding
# NOTE: Extract context. The basic issue is that
# time/location
# here is given at the event level, not at the relation
# level, and so we need to choose which event's context
# we will associate with the relation
def choose_context(context_type):
locs = [
self._events[cause_idx].get(context_type),
self._events[effect_idx].get(context_type)
]
if locs[0]:
return locs[0].strip()
elif locs[1]:
return locs[1].strip()
else:
return None
context = WorldContext()
location = choose_context('Location')
if location:
context.location = RefContext(name=location)
time = choose_context('Time')
if time:
context.time = TimeContext(text=time)
# Overwrite blank context
if not context:
context = None
ev = Evidence(source_api='sofia',
pmid=ref,
annotations=annots,
text=text,
context=context)
stmt = Influence(cause_concept, effect_concept, evidence=[ev])
# Assume unknown polarity on the subject, put the overall
# polarity in the sign of the object
stmt.subj_delta['polarity'] = None
stmt.obj_delta['polarity'] = pol
stmt_list.append(stmt)
return stmt_list