def __create_span(ann_obj, mods, type, start, end, txt_file_path,
projectconf, attributes):
# TODO: Rip this out!
start = int(start)
end = int(end)
# Before we add a new trigger, does it already exist?
found = None
for tb_ann in ann_obj.get_textbounds():
try:
if (tb_ann.start == start and tb_ann.end == end
and tb_ann.type == type):
found = tb_ann
break
except AttributeError:
# Not a trigger then
pass
if found is None:
# Get a new ID
new_id = ann_obj.get_new_id('T') #XXX: Cons
# Get the text span
with open_textfile(txt_file_path, 'r') as txt_file:
text = txt_file.read()[start:end]
#TODO: Data tail should be optional
if '\n' not in text:
ann = TextBoundAnnotationWithText(start, end, new_id, type, text)
ann_obj.add_annotation(ann)
mods.addition(ann)
else:
ann = None
else:
ann = found
if ann is not None:
if projectconf.is_physical_entity_type(type):
# TODO: alert that negation / speculation are ignored if set
event = None
else:
# Create the event also
new_event_id = ann_obj.get_new_id('E') #XXX: Cons
event = EventAnnotation(ann.id, [], unicode(new_event_id), type, '')
ann_obj.add_annotation(event)
mods.addition(event)
else:
# We got a newline in the span, don't take any action
event = None
return ann, event
def _create_argument(ann_obj, projectconf, mods, origin, target, type,
attributes, old_type, old_target):
try:
arg_tup = (type, str(target.id))
# Is this an addition or an update?
if old_type is None and old_target is None:
if arg_tup not in origin.args:
before = str(origin)
origin.add_argument(type, str(target.id))
mods.change(before, origin)
else:
# It already existed as an arg, we were called to do nothing...
pass
else:
# Construct how the old arg would have looked like
old_arg_tup = (type if old_type is None else old_type,
target if old_target is None else old_target)
if old_arg_tup in origin.args and arg_tup not in origin.args:
before = str(origin)
origin.args.remove(old_arg_tup)
origin.add_argument(type, str(target.id))
mods.change(before, origin)
else:
# Collision etc. don't do anything
pass
except AttributeError:
# The annotation did not have args, it was most likely an entity
# thus we need to create a new Event...
new_id = ann_obj.get_new_id('E')
ann = EventAnnotation(
origin.id,
[arg_tup],
new_id,
origin.type,
''
)
ann_obj.add_annotation(ann)
mods.addition(ann)
# No addressing mechanism for arguments at the moment
return None
def __create_span(ann_obj, mods, type, offsets, txt_file_path,
projectconf, attributes):
# For event types, reuse trigger if a matching one exists.
found = None
if projectconf.is_event_type(type):
for tb_ann in ann_obj.get_textbounds():
try:
if (_offsets_equal(tb_ann.spans, offsets)
and tb_ann.type == type):
found = tb_ann
break
except AttributeError:
# Not a trigger then
pass
if found is None:
# Get a new ID
new_id = ann_obj.get_new_id('T') # XXX: Cons
# Get the text span
with open_textfile(txt_file_path, 'r') as txt_file:
text = txt_file.read()
text_span = _text_for_offsets(text, offsets)
# The below code resolves cases where there are newlines in the
# offsets by creating discontinuous annotations for each span
# separated by newlines. For most cases it preserves the offsets.
seg_offsets = []
for o_start, o_end in offsets:
pos = o_start
for text_seg in text_span.split('\n'):
if not text_seg and o_start != o_end:
# Double new-line, skip ahead
pos += 1
continue
start = pos
end = start + len(text_seg)
# For the next iteration the position is after the newline.
pos = end + 1
# Adjust the offsets to compensate for any potential leading
# and trailing whitespace.
start += len(text_seg) - len(text_seg.lstrip())
end -= len(text_seg) - len(text_seg.rstrip())
# If there is any segment left, add it to the offsets.
if start != end:
seg_offsets.append((start, end, ))
# if we're dealing with a null-span
if not seg_offsets:
seg_offsets = offsets
ann_text = DISCONT_SEP.join((text[start:end]
for start, end in seg_offsets))
ann = TextBoundAnnotationWithText(seg_offsets, new_id, type, ann_text)
ann_obj.add_annotation(ann)
mods.addition(ann)
else:
ann = found
if ann is not None:
if projectconf.is_physical_entity_type(type):
# TODO: alert that negation / speculation are ignored if set
event = None
else:
# Create the event also
new_event_id = ann_obj.get_new_id('E') # XXX: Cons
event = EventAnnotation(
ann.id, [], str(new_event_id), type, '')
ann_obj.add_annotation(event)
mods.addition(event)
else:
# We got a newline in the span, don't take any action
event = None
return ann, event
def __create_span(ann_obj, mods, type, offsets, txt_file_path, projectconf,
attributes):
# For event types, reuse trigger if a matching one exists.
found = None
if projectconf.is_event_type(type):
for tb_ann in ann_obj.get_textbounds():
try:
if (_offsets_equal(tb_ann.spans, offsets)
and tb_ann.type == type):
found = tb_ann
break
except AttributeError:
# Not a trigger then
pass
if found is None:
# Get a new ID
new_id = ann_obj.get_new_id('T') #XXX: Cons
# Get the text span
with open_textfile(txt_file_path, 'r') as txt_file:
# TODO discont: use offsets instead (note need for int conversion)
text = _text_for_offsets(txt_file.read(), offsets)
# The below code resolves cases where there are newlines in the
# offsets by creating discontinuous annotations for each span
# separated by newlines. For most cases it preserves the offsets.
seg_offsets = []
for o_start, o_end in offsets:
pos = o_start
for text_seg in text.split('\n'):
if not text_seg and o_start != o_end:
# Double new-line, skip ahead
pos += 1
continue
end = pos + len(text_seg)
seg_offsets.append((pos, end))
# Our current position is after the newline
pos = end + 1
ann = TextBoundAnnotationWithText(
seg_offsets,
new_id,
type,
# Replace any newlines with the discontinuous separator
MUL_NL_REGEX.sub(DISCONT_SEP, text))
ann_obj.add_annotation(ann)
mods.addition(ann)
else:
ann = found
if ann is not None:
if projectconf.is_physical_entity_type(type):
# TODO: alert that negation / speculation are ignored if set
event = None
else:
# Create the event also
new_event_id = ann_obj.get_new_id('E') #XXX: Cons
event = EventAnnotation(ann.id, [], unicode(new_event_id), type,
'')
ann_obj.add_annotation(event)
mods.addition(event)
else:
# We got a newline in the span, don't take any action
event = None
return ann, event