def tag_text_with_annotations(self, original_text, print_line_numbers=False):
"""
mark the string original_text with object's annotations
original_text: string with the text to be marked
print_line_numbers: add line numbers to the text
:return output_text the text in string original_text with added marks
footnotes the list of techniques in the text
legend description of the marks added
"""
if Articles_annotations.techniques is None:
if ans.Annotation.propaganda_techniques is None:
Articles_annotations.techniques = Propaganda_Techniques()
else:
Articles_annotations.techniques = ans.Annotation.propaganda_techniques
self.get_markers_from_spans()
output_text, curr_output_text_index, self.curr_marker = ("", 0, 0)
techniques_found = set()
row_counter = 1
#print(self.markers)
annotations_stack = [] # to handle overlapping annotations when assigning color background
while curr_output_text_index < len(original_text):
if self.curr_marker >= len(self.markers): # done marking text, need to flush the remaining content of <original_text> into <output_text>
output_text += original_text[curr_output_text_index:]
curr_output_text_index = len(original_text)
else: # more markers have to be added to the content string
if self.marker_position() <= curr_output_text_index: # it is time to add a marker
techniques_index = self.techniques.indexOf(self.marker_label())
techniques_found.add(techniques_index)
if self.is_starting_marker():
output_text += self.start_annotation_marker_function(annotations_stack, self.curr_marker, row_counter)
annotations_stack.append(self.marker_annotation())
else:
output_text += self.end_annotation_marker_function(annotations_stack, self.curr_marker, row_counter)
annotations_stack.remove(self.marker_annotation())
self.curr_marker += 1
else: # flush string content up to the next marker
text_to_be_added = original_text[curr_output_text_index:self.marker_position()]
row_counter += text_to_be_added.count('\n')
output_text += text_to_be_added
curr_output_text_index = self.marker_position()
final_text = ""
for row_counter, line in enumerate(output_text.split("\n"), 1):
final_text += self.add_sentence_marker(line, row_counter)
footnotes = "\n<div>List of techniques found in the article</div>\n\n"
for technique_index in sorted(techniques_found):
footnotes += "<div>%d: %s</div>\n" % (technique_index, self.techniques.get_technique(technique_index))
return final_text, footnotes
import src.annotation_w_o_label as anwol
from src.propaganda_techniques import Propaganda_Techniques
import logging.handlers
__author__ = "Giovanni Da San Martino"
__copyright__ = "Copyright 2019"
__credits__ = ["Giovanni Da San Martino"]
__license__ = "GPL"
__version__ = "0.1"
__maintainer__ = "Giovanni Da San Martino"
__email__ = "*****@*****.**"
__status__ = "Beta"
logger = logging.getLogger("propaganda_scorer")
techniques = Propaganda_Techniques()
class Articles_annotations(object):
"""
Class for handling annotations for one article.
Articles_annotations is composed of an article id
and a list of Annotation objects.
"""
start_annotation_effect = ""
end_annotation_effect = ""
start_annotation_str = "<span -"
end_annotation_str = "- span>"
annotation_background_color = ""
def mark_text(self, original_text, print_line_numbers=False):
"""
mark the string original_text with object's annotations
original_text: string with the text to be marked
print_line_numbers: add line numbers to the text
:return output_text the text in string original_text with added marks
footnotes the list of techniques in the text
legend description of the marks added
"""
self.get_markers_from_spans()
if Articles_annotations.techniques is None:
if ans.Annotation.propaganda_techniques is None:
Articles_annotations.techniques = Propaganda_Techniques()
else:
Articles_annotations.techniques = ans.Annotation.propaganda_techniques
output_text, curr_output_text_index, self.curr_marker = ("", 0, 0)
footnotes = "List of techniques found in the article\n\n"
techniques_found = set()
annotations_stack = (
[]
) # to handle overlapping annotations when assigning color background
while curr_output_text_index < len(original_text):
if self.curr_marker >= len(self.markers):
output_text += original_text[curr_output_text_index:]
curr_output_text_index = len(original_text)
else:
if self.marker_position() <= curr_output_text_index:
if self.is_starting_marker():
output_text += (self.start_annotation_effect +
self.start_annotation_str)
annotations_stack.append(self.marker_annotation())
else:
output_text += "%s%s%s" % (
self.end_annotation_effect,
"" if len(annotations_stack) > 1 else " ",
self.start_annotation_effect,
)
techniques_index = Articles_annotations.techniques.indexOf(
self.marker_label())
output_text += str(techniques_index)
techniques_found.add(techniques_index)
if self.is_ending_marker():
output_text += (self.end_annotation_str +
self.end_annotation_effect)
annotations_stack.remove(self.marker_annotation())
if len(annotations_stack) > 0:
output_text += self.annotation_background_color
else:
output_text += (self.end_annotation_effect + " " +
self.annotation_background_color)
self.curr_marker += 1
else:
output_text += original_text[curr_output_text_index:self.
marker_position()]
curr_output_text_index = self.marker_position()
if print_line_numbers:
indices, char_index = ([], 0)
for line in original_text.split("\n"):
indices.append(char_index)
char_index += len(line) + 1
# output_text = "\n".join(["%d (%d) %s"%(i, x[0], x[1])
output_text = "\n".join([
"%d %s" % (i, x[1])
for i, x in enumerate(zip(indices, output_text.split("\n")), 1)
])
legend = (
"---\n%sHighlighted text%s: any propagandistic fragment\n%s%si%s: start of the i-th technique"
"\n%si%s%s: end of the i-th technque\n---" % (
self.annotation_background_color,
self.end_annotation_effect,
self.start_annotation_effect,
self.start_annotation_str,
self.end_annotation_effect,
self.start_annotation_effect,
self.end_annotation_str,
self.end_annotation_effect,
))
for technique_index in sorted(techniques_found):
footnotes += "%d: %s\n" % (
technique_index,
Articles_annotations.techniques.get_technique(technique_index),
)
return output_text, footnotes, legend